JP7639205B1 - Shelf allocation proposal system, shelf allocation proposal method, and program - Google Patents

Abstract

[Problem] To improve the accuracy of shelf allocation proposals. [Solution] A learning model storage unit (201) of a shelf allocation proposal system (1) stores a learned learning model (M) that has learned the relationship between training product information on training products for training and training shelf allocation information on the shelf allocation of the training products in a training store for training. An estimated product information acquisition unit (202) acquires estimated product information on estimated products handled in an estimated store whose shelf allocation is estimated. A proposal unit (203) makes a proposal regarding the shelf allocation of the estimated product in the estimated store based on the learning model (M) and the estimated product information. [Selected Figure] Figure 3

Description

This disclosure relates to a shelf layout proposal system, a shelf layout proposal method, and a program.

Traditionally, in stores where products are arranged on shelves, a task known as shelf allocation is carried out in which a store or manufacturer's staff member decides the arrangement of products. For example, Patent Literature 1 describes a technology that generates an ideal shelf allocation pattern, which is an ideal shelf allocation pattern, based on the shelf allocation patterns of each of multiple shelves in a store, calculates the difference between the ideal shelf allocation pattern and an actual shelf allocation pattern, which is the shelf allocation pattern applied to the proposed shelf, and corrects the ideal shelf allocation pattern based on market trends and the difference to generate a final proposed shelf allocation pattern.

For example, Patent Document 2 describes a technology that optimizes the layout of shelves at a logistics site based on the distance between parts arranged on the shelves. Patent Document 3 describes a technology that predicts product sales based on relationship information that represents the relationship between the arrangement of products on store shelves and product sales, and proposes shelf allocation that is expected to produce the highest sales. Patent Document 3 also describes recognizing products by using images of shelves photographed by a camera as learning data.

JP 2004-227272 A JP 2015-079318 A JP 2023-030023 A

However, the technology of Patent Document 1 generates an ideal shelf allocation pattern within a range of predetermined shelf allocation patterns, so it cannot accommodate stores that do not have a shelf allocation pattern defined, and the accuracy of shelf allocation proposals cannot be sufficiently improved. The technology of Patent Document 2 is based on shelves at the logistics site, so it is difficult to apply it to shelf allocations at stores that sell products. The technology of Patent Document 3 proposes shelf allocations within a range of predetermined relationship information, so it cannot accommodate stores that do not have relationship information defined, and the accuracy of shelf allocation proposals cannot be sufficiently improved.

One of the objectives of this disclosure is to improve the accuracy of shelf allocation proposals.

The shelf allocation proposal system according to the present disclosure includes a learning model storage unit that stores a learned learning model in which the relationship between training product information on training products for training and training shelf allocation information on the shelf allocation of the training products in a training store for training is learned, an estimated product information acquisition unit that acquires estimated product information on estimated products handled in an estimated store whose shelf allocation is estimated, and a proposal unit that makes a proposal for shelf allocation of the estimated products in the estimated store based on the learning model and the estimated product information.

This disclosure can improve the accuracy of shelf allocation proposals.

FIG. 2 is a diagram illustrating an example of a hardware configuration of the shelf allocation proposal system. FIG. 13 is a diagram illustrating an example of a shelf layout proposal service that the shelf layout proposal system provides to a person in charge. FIG. 2 is a diagram illustrating an example of functions realized by the shelf allocation proposal system. FIG. 2 is a diagram illustrating an example of a training database. FIG. 13 is a diagram illustrating an example of an estimation database. FIG. 13 is a diagram showing an example of a shelf allocation proposal screen. FIG. 2 is a diagram illustrating an example of a process executed in the shelf allocation proposal system. FIG. 13 is a diagram illustrating an example of a function realized in a modified example.

[1. Hardware configuration of shelf layout proposal system]
An example of an embodiment of a shelf allocation proposal system, a shelf allocation proposal method, and a program according to the present disclosure will be described. FIG. 1 is a diagram showing an example of a hardware configuration of the shelf allocation proposal system. For example, the shelf allocation proposal system 1 includes a learning terminal 10, a server 20, and a person in charge terminal 30. Each of the learning terminal 10, the server 20, and the person in charge terminal 30 is connected to a network N such as the Internet or a LAN. Although FIG. 1 shows one each of the learning terminal 10, the server 20, and the person in charge terminal 30, there may be a plurality of at least one of the learning terminal 10, the server 20, and the person in charge terminal 30.

The learning terminal 10 is a computer that learns the learning model described below. For example, the learning terminal 10 is a personal computer, a server computer, a tablet, or a smartphone. For example, the learning terminal 10 includes a control unit 11, a memory unit 12, a communication unit 13, an operation unit 14, and a display unit 15. The control unit 11 includes at least one processor. The memory unit 12 includes at least one of a volatile memory such as a RAM and a non-volatile memory such as a flash memory. The communication unit 13 includes at least one of a communication interface for wired communication and a communication interface for wireless communication. The operation unit 14 is an input device such as a touch panel or a mouse. The display unit 15 is a display such as a liquid crystal or organic EL.

The server 20 is a server computer of a service provider that provides a shelf allocation proposal service. The shelf allocation proposal service is a service that proposes shelf allocation to a person in charge of shelf allocation. The person in charge is a person in charge of shelf allocation work. The person in charge may be any person. For example, the person in charge may be a store clerk or a manufacturer that produces the products. For example, the server 20 includes a control unit 21, a memory unit 22, and a communication unit 23. The hardware configurations of the control unit 21, the memory unit 22, and the communication unit 23 may be the same as those of the control unit 11, the memory unit 12, and the communication unit 13, respectively.

The staff terminal 30 is the staff's computer. For example, the staff terminal 30 is a personal computer, a POS terminal, a tablet, or a smartphone. The staff terminal 30 includes a control unit 31, a memory unit 32, a communication unit 33, an operation unit 34, and a display unit 35. The hardware configurations of the control unit 31, the memory unit 32, the communication unit 33, the operation unit 34, and the display unit 35 may be similar to those of the control unit 11, the memory unit 12, the communication unit 13, the operation unit 14, and the display unit 15, respectively.

The programs stored in the storage units 12, 22, 32 may be supplied to the learning terminal 10, the server 20, or the staff terminal 30 via the network N. At least one of a reading unit (e.g., a memory card slot) that reads a computer-readable information storage medium and an input/output unit (e.g., a USB port) for inputting and outputting data to and from an external device may be included in the learning terminal 10, the server 20, or the staff terminal 30. For example, a program stored in an information storage medium may be supplied to the learning terminal 10, the server 20, or the staff terminal 30 via at least one of the reading unit and the input/output unit.

Furthermore, the shelf allocation proposal system 1 may include at least one computer. The computers included in the shelf allocation proposal system 1 are not limited to the example in FIG. 1. For example, the shelf allocation proposal system 1 may include only the learning terminal 10 and the server 20. In this case, the person in charge terminal 30 exists outside the shelf allocation proposal system 1. The shelf allocation proposal system 1 may include only the server 20. In this case, the learning terminal 10 and the person in charge terminal 30 exist outside the shelf allocation proposal system 1. For example, the shelf allocation proposal system 1 may include the server 20 and other computers not shown in FIG. 1.

[2. Overview of the shelf layout proposal system]
In this embodiment, an example is taken of a case where a person working for a manufacturer that produces products handled in a store performs shelf allocation. The person in charge may perform shelf allocation only for products produced by the manufacturer he or she works for, but in this embodiment, the person in charge also performs shelf allocation for products produced by other manufacturers. The person in charge performs optimal shelf allocation by comprehensively considering various factors such as the tastes of customers visiting the store, trends in the world, and the season. The shelf allocation proposal system 1 provides the person in charge with a shelf allocation proposal service to reduce the burden on the person in charge.

Figure 2 is a diagram showing an example of a shelf allocation proposal service provided to a person in charge by the shelf allocation proposal system 1. As shown in Figure 2, the person in charge allocates at least a portion of the multiple shelves arranged in the store SP. In this embodiment, the shelf allocation proposal system 1 proposes an optimal shelf allocation to the person in charge based on a trained learning model M. The learning model M is a model trained by a machine learning method. The machine learning method may be any of various known methods. The method for training the learning model M may also be any of various known methods.

There are various definitions of machine learning, and the machine learning of this embodiment may be machine learning in various known definitions. For example, the machine learning of this embodiment includes deep learning. The learning model M trained by the machine learning method of this embodiment includes AI. For example, the learning model M may be generative AI. For example, the learning model M may be a neural network, a boosting model (decision tree model), a support vector machine, a large-scale language model, or other models.

For example, multiple pieces of training data are learned in the learning model M. The training data includes an input portion that is input to the learning model M during learning, and an output portion that indicates the content that should be output from the learning model M during learning. The output portion can also be said to be the correct answer during learning. The format of the input portion of the training data is basically the same as the format of the estimated data that is input to the learning model M during estimation. The format of the output portion of the training data is basically the same as the format of the data that is output to the learning model M during estimation.

In this embodiment, the input portion of the training data includes training product information, training shelf information, and training purchase information. The input portion of the training data is not limited to the example of this embodiment. The input portion of the training data only needs to include training product information. For example, the input portion of the training data may be only training product information, without including training shelf information and training purchase information. The input portion of the training data may be only training product information and training shelf information, without including training purchase information. The input portion of the training data may be only training product information and training shelf information, without including training shelf information.

Training product information is information about training products, which are products for training. Training products are not limited to products that exist in reality, and may be fictitious products. Training product information indicates the characteristics of the training product. For example, the training product information indicates product identification information such as the product name, content, type, price, size, design, taste, texture, JAN code, or other characteristics of the training product. The training product information may indicate multiple characteristics of the training product. In this embodiment, an example is given in which the training product information indicates the characteristics of each of multiple training products, but the training product information may only indicate the characteristics of one training product. All training products may be placed, or some training products may not be placed. The training product information may indicate training products that have been decided to be placed on the shelf, or may indicate training products that have not been decided to be placed on the shelf (training products that are candidates for placement on the shelf).

Training shelf information is information about a training shelf, which is a shelf used for training. A training shelf is a shelf arranged in a training store, which is a store used for training. The training store may be a real store or a fictitious store. Training shelf information indicates characteristics of a training shelf. For example, training shelf information indicates the size, capacity, number of levels, type, design, material, location in the store, or other characteristics of a training shelf. Training shelf information may indicate multiple characteristics related to a training shelf. In this embodiment, an example is given in which the training shelf information indicates the characteristics of one training shelf, but the training shelf information may only indicate the characteristics of multiple training shelves. Furthermore, training shelf information may only indicate the characteristics of a portion of the space in the training shelf.

The training purchase information is information about purchases at a training store. For example, the training purchase information may be information about purchases of training products placed on training shelves, or information about purchases of training products not placed on training shelves. The training purchase information indicates the characteristics of purchases at a training store. For example, the training purchase information indicates the type, price, quantity, total amount, or other characteristics of the training products purchased at the training store. The training purchase information may indicate changes over time in these. The training purchase information may indicate the characteristics of training customers who have purchased training products. For example, the training purchase information may indicate the gender, age group, preferences, behavior, or other characteristics of the training customers. In this embodiment, an example is given of the case where the training purchase information indicates the characteristics of purchases at one training store, but the training purchase information may also indicate the characteristics of purchases at each of multiple training stores. The training purchase information may include information acquired from a POS terminal at the training store.

In this embodiment, the output portion of the training data is training shelf allocation information. For example, the training shelf allocation information includes training promotion/demotion information and training substitute product information. The output portion of the training data is not limited to the example of this embodiment. For example, the output portion of the training data may be only either the training promotion/demotion information or the training substitute product information. The output portion of the training data may be information other than the training promotion/demotion information and the training substitute product information.

Training promotion/demotion information is information about the promotion and demotion of training products. Promotion is an increase in importance in the shelf allocation. For example, promotion is a change from a state where a training product is not placed on a training shelf to a state where it is placed on the training shelf, a change in the position of the training product from the top or bottom to the middle of the training shelf, a change in the position of the training product from the left end or right end of the training shelf to the center, an increase in the number of training products (e.g., the number of rows), or other changes. Demotion is a decrease in importance in the shelf allocation. For example, promotion is a change from a state where a training product is placed on a training shelf to a state where it is not placed on the training shelf, a change in the position of the training product from the middle to the top or bottom of the training shelf, a change in the position of the training product from the center to the left end or right end of the training shelf, a decrease in the number of training products (e.g., the number of rows), or other changes. In this embodiment, an example is given of a case where the training promotion/demotion information is information about the promotion and demotion of a training product indicated by the training product information. For this reason, an example will be given in which the training promotion/demotion information is information regarding the promotion and demotion of each of multiple training products. The training promotion/demotion information may be information regarding the promotion and demotion of a training product that is not indicated in the training product information. The training promotion/demotion information may be information regarding the promotion and demotion of a single training product.

Training substitute product information is information about training substitute products that substitute for training products. Training substitute products can also be said to be training products that are placed on the training shelf in place of training products placed on the training shelf. Training substitute product information may be information that can identify the training substitute product (e.g., product name or JAN code), or may be information that indicates the characteristics of the training substitute product. For example, the training substitute product information indicates product identification information such as the product name, content, type, price, size, design, taste, texture, JAN code, or other characteristics of the training substitute product. The training substitute product information may indicate multiple characteristics of the training substitute product. In this embodiment, an example is given in which the training substitute product information indicates the characteristics of each of multiple training substitute products, but the training substitute product information may only indicate the characteristics of one training substitute product.

Note that the training shelf allocation information is not limited to training promotion/demotion information and training substitute product information. The training shelf allocation information may be any information related to shelf allocation that is the correct answer during learning. For example, the training shelf allocation information may indicate a list of training products that should be placed on the training shelf. Furthermore, the training shelf allocation information may indicate a list of training products that should be placed in each space on the training shelf (e.g., each of the spaces on the upper, middle, and lower levels). The training shelf allocation information may indicate a list of training products that should not be placed on the training shelf. For example, the training shelf allocation information may indicate a score (probability) indicating whether a training product should be placed on the training shelf.

In this embodiment, the above-mentioned training data is learned by the learning model M. For example, the learning terminal 10 causes the learning model M to learn the training data. The learning terminal 10 may cause the learning model M to learn the training data by including information on the arrangement of training products on the training shelves and the corresponding sales of the products in the training data. The learning terminal 10 may train the learning model M based on the training data in which the arrangement of the training products is labeled according to whether the sales are above or below some benchmark sales.

For example, the learning terminal 10 uploads the trained learning model M to the server 20. The server 20 receives the trained learning model M and records it in the memory unit 22. When a staff member considers shelf allocation for the store he or she is in charge of, the staff member operates the staff member terminal 30 to log in to the shelf allocation proposal service and uses the trained learning model M recorded in the server 20.

For example, the person in charge operates the person in charge terminal 30 to input estimated data including estimated product information, estimated shelf information, and estimated purchase information into the trained learning model M. The estimated data is assumed to be pre-stored in the person in charge terminal 30. The format of the estimated data is basically the same as the format of the input portion of the training data. Therefore, the format of the estimated product information is basically the same as the format of the training product information. The format of the estimated shelf information is basically the same as the format of the training shelf information. The format of the estimated purchase information is basically the same as the format of the training purchase information.

The estimated product information is information about an estimated product, which is a product for estimation. The estimated product information indicates the characteristics of the estimated product. For example, the estimated product information indicates product identification information such as the product name, content, type, price, size, design, taste, texture, JAN code, or other characteristics of the estimated product. The estimated product information may indicate multiple characteristics of the estimated product. In this embodiment, the estimated product information indicates the characteristics of each of multiple estimated products, but the estimated product information may only indicate the characteristics of one estimated product. All estimated products may be placed, or some estimated products may not be placed. The estimated product information may indicate estimated products that have been decided to be placed on the shelf, or may indicate estimated products that have not been decided to be placed on the shelf (estimated products that are candidates for placement on the shelf).

The estimated shelf information is information about an estimated shelf, which is an estimated shelf. The estimated shelf is a shelf arranged in an estimated store, which is an estimated store. The estimated store may be an actual store or a fictitious store. The estimated shelf information indicates the characteristics of the estimated shelf. For example, the estimated shelf information indicates the size, capacity, number of shelves, type, design, material, location in the store, function such as refrigerated/non-refrigerated, or other characteristics of the estimated shelf. The estimated shelf information may indicate multiple characteristics related to the estimated shelf. In this embodiment, an example is given in which the estimated shelf information indicates the characteristics of one estimated shelf, but the estimated shelf information may only indicate the characteristics of multiple estimated shelves. Furthermore, the estimated shelf information may only indicate the characteristics of a portion of the space in the estimated shelf.

The estimated purchase information is information about purchases in an estimated store. For example, the estimated purchase information may be information about purchases of estimated products placed on estimated shelves, or information about purchases of estimated products not placed on estimated shelves. The estimated purchase information indicates the characteristics of purchases in an estimated store. For example, the estimated purchase information indicates the type, price, quantity, total amount, or other characteristics of the estimated products purchased in the estimated store. The estimated purchase information may indicate changes over time. The estimated purchase information may indicate the characteristics of an estimated customer who purchased an estimated product. For example, the estimated purchase information may indicate the gender, age group, or other characteristics of the estimated customer. In this embodiment, the estimated purchase information indicates the characteristics of purchases in one estimated store, but the estimated purchase information may indicate the characteristics of purchases in each of multiple estimated stores. The estimated purchase information may indicate the characteristics of past purchases in the estimated store, or may indicate the characteristics of real-time purchases in the estimated store. The estimated purchase information may include information acquired from a POS terminal in the estimated store.

For example, when estimated data is input, the learning model M calculates an embedded representation (feature amount) of the estimated data based on parameters adjusted by learning. In this embodiment, the embedded representation is a multidimensional vector as an example. The embedded representation may be in a format other than a vector. For example, the embedded representation may be in an array format, a matrix format, multiple numerical values, a single numerical value, or another format. The learning model M outputs estimated shelf allocation information regarding shelf allocation that is estimated to be appropriate based on the embedded representation.

For example, the estimated shelf allocation information includes estimated promotion/demotion information and estimated substitute product information. The output from the learning model M is not limited to the example of this embodiment. For example, the output from the learning model M may be only either the estimated promotion/demotion information or the estimated substitute product information. The output from the learning model M may be information other than the estimated promotion/demotion information and the estimated substitute product information.

The estimated promotion/demotion information is information regarding the promotion and demotion of an estimated product. In this embodiment, an example is given in which the estimated promotion/demotion information is information regarding the promotion and demotion of an estimated product indicated by the estimated product information. Therefore, an example is given in which the estimated promotion/demotion information is information regarding the promotion and demotion of each of a plurality of estimated products. The estimated promotion/demotion information may be information regarding the promotion and demotion of an estimated product not indicated in the estimated product information. The estimated promotion/demotion information may be information regarding the promotion and demotion of a single estimated product.

The estimated substitute product information is information about an estimated substitute product that is a substitute for the estimated product. The estimated substitute product can also be said to be an estimated product that is placed on the estimated shelf in place of the estimated product placed on the estimated shelf. The estimated substitute product information may be information that can identify the estimated substitute product (e.g., product name or JAN code), or may be information that indicates the characteristics of the estimated substitute product. For example, the estimated substitute product information indicates product identification information such as the product name, content, type, price, size, design, taste, texture, JAN code, etc. of the estimated substitute product, or other characteristics. The estimated substitute product information may indicate multiple characteristics of the estimated substitute product. In this embodiment, an example is given in which the estimated substitute product information indicates the characteristics of each of multiple estimated substitute products, but the estimated substitute product information may only indicate the characteristics of one estimated substitute product.

The estimated shelf allocation information is not limited to estimated promotion/demotion information and estimated substitute product information. The estimated shelf allocation information may be any information related to shelf allocation that the learning model M estimates to be appropriate. For example, the estimated shelf allocation information may indicate a list of estimated products that should be placed on the estimated shelf. Furthermore, the estimated shelf allocation information may indicate, for each space on the estimated shelf (e.g., each space on the upper, middle, and lower levels), a list of estimated products that should be placed in that space. The estimated shelf allocation information may indicate a list of estimated products that should not be placed on the estimated shelf. For example, the estimated shelf allocation information may indicate a score (probability) indicating whether the estimated product should be placed on the estimated shelf.

For example, the person in charge checks the estimated shelf allocation information output by the learning model M and considers shelf allocation. The person in charge may complete shelf allocation using the estimated shelf allocation information as is, or may complete shelf allocation by modifying the estimated shelf allocation information. The person in charge may perform shelf allocation by themselves by referring to the estimated shelf allocation information. As described above, the shelf allocation proposal system 1 of this embodiment is capable of supporting shelf allocation performed by the person in charge based on the learning model M in which training data such as that shown in FIG. 2 has been learned. The details of the shelf allocation proposal system 1 will be described below.

[3. Functions realized by the shelf layout proposal system]
FIG. 3 is a diagram showing an example of functions realized by the shelf allocation proposal system 1.

[3-1. Functions realized on the learning device]
For example, the learning terminal 10 includes a data storage unit 100, a learning model storage unit 101, and a learning unit 102. Each of the data storage unit 100 and the learning model storage unit 101 is realized by the storage unit 12. The learning unit 102 is realized by the control unit 11.

[Data storage unit]
The data storage unit 100 stores data necessary for learning the learning model M. For example, the data storage unit 100 stores a training database DB1. Note that the data stored in the data storage unit 100 is not limited to the training database DB1. The data storage unit 100 only needs to store data necessary for learning the learning model M.

Figure 4 is a diagram showing an example of the training database DB1. For example, multiple training data are stored in the training database DB1. Examples of individual training data are as described with reference to Figure 2. In this embodiment, the service provider of the shelf allocation proposal service prepares the training data. The training data may not be prepared manually, but may be created automatically or semi-automatically by a training data creation tool. The training data creation tool may be a publicly known tool.

[Learning model storage unit]
The learning model storage unit 101 stores the learning model M. The learning model M includes a program indicating various processes required for estimating an appropriate shelf allocation, and parameters referenced by the program. The program and parameters of the learning model M may be similar to programs and parameters used in known machine learning techniques. For example, the program of the learning model M includes program code indicating calculation of an embedded representation, and program code indicating processing of an output according to the embedded representation. The parameters may be weights, biases, or other parameters.

For example, the learning model storage unit 101 stores a learning model M with parameters of initial values. When learning by the learning unit 102 is completed, the learning model storage unit 101 stores the learned learning model M. The learning model M with parameters of initial values may be overwritten with the learned learning model M, or may be left in the learning model storage unit 101 separately from the learned learning model M. A learning model M that has undergone some pre-learning, rather than a learning model M with parameters of initial values, may be stored in the learning model storage unit 101. The learning model M that has undergone pre-learning may be tuned by the learning unit 102.

[Learning Department]
The learning unit 102 learns the learning model M based on the training data stored in the training database DB1. The learning of the learning model M is the adjustment of the parameters of the learning model M. For example, the learning unit 102 learns the learning model M so that when an input portion of the training data is input to the learning model M, an output portion of the training data is output. The learning algorithm may be the same as an algorithm adopted in a known machine learning method. For example, the learning unit 102 may cause the learning model M to learn the training data based on an algorithm such as a gradient descent method or an error backpropagation method. The function such as a loss function used by the learning unit 102 may also be a function adopted in a known machine learning method. When the learning of the learning model M is completed, the learning unit 102 records the learned learning model M in the learning model storage unit 101. The learning unit 102 uploads the learned learning model M to the server 20.

In this embodiment, the input portion of the training data includes training product information, training shelf information, and training purchase information. The output portion of the training data includes training promotion/demotion information and training substitute product information as training shelf allocation information. A correlation exists between the input portion of the training data and the output portion of the training data. The learning model M learns the correlation between them by learning the training data. As a result, when the proposal unit 205 described below inputs estimated data to the learning model M, the learning model M can output estimated promotion/demotion information and estimated substitute product information as estimated shelf allocation information according to the estimated data, based on the correlation it has learned (i.e., by processing based on parameters adjusted by learning).

For example, training products with a certain product name may be better placed on the training shelf. On the other hand, training products with other product names may not be better placed on the training shelf. Furthermore, training products with a certain product name may be better placed near the center of the training shelf. On the other hand, training products with other product names may be better placed near the edge of the training shelf. In this case, there is a correlation between the product name indicated by the training product information and the content of the training shelf allocation information (for example, at least one of the promotion/demotion indicated by the training promotion/demotion information and the substitute product indicated by the training substitute product information). The learning model M learns the correlation between them by learning the training data. As a result, the learning model M can output estimated shelf allocation information indicating content that is correlated with the product name indicated by the estimated product information included in the estimation data (for example, at least one of the promotion/demotion indicated by the estimated promotion/demotion information indicating the promotion/demotion according to the product name and the estimated substitute product information indicating the substitute product according to the product name). The same applies when the training product information and the estimated product information each indicate content other than the product name. For example, the learning model M can learn correlations between the contents of the training planogram information and product identification information such as the product contents, type, price, size, design, taste, texture, and JAN code, or other characteristics of the product. The learning model M outputs according to the correlations it has learned.

For example, a training product that is better placed on a training shelf of a certain size may differ from a training product that is better placed on a training shelf of another size. Furthermore, a training product that is better placed at a specific position (e.g., near the center or edge) of a training shelf of a certain size may differ from a training product that is better placed at the same position of a training shelf of another size. In this case, there is a correlation between the size of the training shelf indicated by the training shelf information and the content of the training shelf allocation information (e.g., at least one of the promotion/demotion indicated by the training promotion/demotion information and the substitute product indicated by the training substitute product information). The learning model M learns the correlation between them by learning the training data. As a result, the learning model M can output estimated shelf allocation information that indicates content that is correlated with the size of the estimated shelf indicated by the estimated shelf information included in the estimation data (e.g., at least one of the promotion/demotion indicated by the estimated promotion/demotion information indicating the promotion/demotion according to the size of the estimated shelf and the estimated substitute product information indicating the substitute product according to the size of the estimated shelf). The same applies when the training shelf information and the estimated shelf information each indicate content other than size. For example, the learning model M can learn correlations between the contents of the training planogram information and the capacity, number of shelves, type, design, material, location in the store, or other features that have a correlation with the contents of the training planogram information. The learning model M outputs according to the correlations it has learned.

For example, a training product that should be placed on a training shelf in a training store where a certain type of training product is likely to be purchased may differ from a training product that should be placed on a training shelf in a training store where another type of training product is likely to be purchased. Furthermore, a training product that should be placed in a specific position (e.g., near the center or edge) of a training shelf in a training store where a certain type of training product is likely to be purchased may differ from a training product that should be placed in the same position of a training shelf in a training store where another type of training product is likely to be purchased. In this case, there is a correlation between the type of training product indicated by the training purchase information and the content of the training shelf allocation information (e.g., at least one of the promotion/demotion indicated by the training promotion/demotion information and the substitute product indicated by the training substitute product information). The learning model M learns the correlation between them by learning the training data. This allows the learning model M to output estimated shelf allocation information indicating content that is correlated with the type of estimated product indicated by the estimated purchasing information included in the estimated data (for example, at least one of promotion/demotion indicated by estimated promotion/demotion information indicating promotion/demotion according to the type of training product that is likely to be purchased, and estimated substitute product information indicating substitute products according to the type of training product that is likely to be purchased). The same applies when the training purchasing information and the estimated purchasing information each indicate content other than type. For example, the learning model M can learn the correlation between the price, quantity, total amount, gender, age group, tastes, behavior, or other characteristics of the training customer that are correlated with the content of the training purchasing information. The learning model M outputs according to the correlation that it has learned.

[3-2. Functions realized by the server]
For example, the server 20 includes a data storage unit 200, a learning model storage unit 201, an estimated product information acquisition unit 202, an estimated shelf information acquisition unit 203, an estimated purchase information acquisition unit 204, and a proposal unit 205. Each of the data storage unit 200 and the learning model storage unit 201 is realized by the storage unit 22. Each of the estimated product information acquisition unit 202, the estimated shelf information acquisition unit 203, the estimated purchase information acquisition unit 204, and the proposal unit 205 is realized by the control unit 21.

[Data storage unit]
The data storage unit 200 stores data necessary for proposing shelf allocation. For example, the data storage unit 200 stores an estimation database DB2. Note that the data stored in the data storage unit 200 is not limited to the estimation database DB2. The data storage unit 200 only needs to store data necessary for proposing shelf allocation. For example, the data storage unit 200 may store a user database in which various information on users (e.g., personnel at a manufacturer) who use the shelf allocation proposal service is stored. The user database stores information such as a user ID and a password for users to log in to the shelf allocation proposal service.

Figure 5 is a diagram showing an example of the estimation database DB2. For example, the estimation database DB2 stores store identification information for identifying a store and estimated data for that store. The store identification information is information that can identify a store in the shelf allocation proposal service. For example, the store identification information is a store ID, store name, address, or other information assigned to the store. The estimated data for the store includes estimated product information for products handled at the store, estimated shelf information for shelves located at the store, and estimated purchase information for purchases at the store.

In this embodiment, the person in charge of shelf allocation operates the person in charge terminal 30 to register the estimated product information and estimated shelf information in the estimated database DB2. For example, when the person in charge operates the person in charge terminal 30 to log in to the shelf allocation proposal service, the person in charge inputs the estimated product information of the estimated product handled at the store for which the person in charge of shelf allocation is responsible. The person in charge inputs the estimated product's product identification information such as the product name, content, type, price, size, design, taste, texture, JAN code, or other characteristics. The person in charge may input information about each of multiple estimated products. The person in charge may input the store ID of the store for which the person in charge of shelf allocation is responsible. The person in charge terminal 30 uploads to the server 20 the store ID of the store for which the person in charge is responsible for shelf allocation and the estimated product information of the estimated product input by the person in charge.

For example, the server 20 stores in the estimation database DB2 the store ID of the store for which the person in charge is in charge of shelf allocation, and the estimated product information received from the person in charge terminal 30, in association with each other. The estimated product information may be entered by a store clerk, rather than by the person in charge. The estimated product information may also be stored in advance in the person in charge terminal 30, the store terminal, other terminals, or an information storage medium, rather than being entered manually. In this case, the store ID indicating which store the estimated product information corresponds to is also stored in advance. The server 20 may acquire the estimated product information stored in advance in these locations, and store the estimated product information in association with the store ID of the store that handles the estimated product, in the estimation database DB2.

For example, when a staff member operates the staff member terminal 30 to log in to the shelf allocation proposal service, the staff member inputs estimated shelf information of the estimated shelves located in the store for which the staff member is responsible for shelf allocation. The staff member inputs the size, capacity, number of shelves, type, design, material, location in the store, function such as refrigerated/non-refrigerated, or other features of the estimated shelves located in the store for which the staff member is responsible for shelf allocation. The staff member terminal 30 uploads to the server 20 the store ID of the store for which the staff member is responsible for shelf allocation and the estimated shelf information of the estimated shelves input by the staff member.

For example, server 20 stores in estimation database DB2 the store ID of the store for which the person in charge is in charge of shelf allocation, and the estimated shelf information received from person in charge terminal 30, in association with each other. Note that the estimated shelf information may be entered by a store clerk, rather than by the person in charge. Also, the estimated shelf information may not be entered manually, but may be stored in advance in person in charge terminal 30, the store terminal, other terminals, or information storage media. In this case, the store ID indicating which store the estimated shelf information corresponds to is also stored in advance. Server 20 may acquire the estimated shelf information stored in advance in these, and store the estimated shelf information in association with the store ID of the store that handles the estimated shelf in estimation database DB2.

For example, when a staff member operates the staff member terminal 30 to log in to the shelf allocation proposal service, the staff member inputs estimated purchase information for the store for which the staff member is responsible for shelf allocation. The staff member inputs the type, price, quantity, total amount, or other characteristics of the estimated product purchased at the estimated store for which the staff member is responsible for shelf allocation. The staff member terminal 30 uploads to the server 20 the store ID of the store for which the staff member is responsible for shelf allocation and the estimated purchase information of the estimated store input by the staff member.

For example, the server 20 stores in the estimation database DB2 the store ID of the store for which the person in charge is in charge of shelf allocation, and the estimated purchase information received from the person in charge terminal 30, in association with each other. The estimated purchase information may be entered by a store clerk, rather than by the person in charge. The estimated purchase information may also be stored in advance in the person in charge terminal 30, the store terminal, other terminals, or an information storage medium, rather than being entered manually. In this case, the store ID indicating which store the estimated purchase information corresponds to is also stored in advance. The server 20 may acquire the estimated purchase information stored in advance and store the estimated purchase information in the estimation database DB2 in association with the store ID of the store that handles the estimated product.

[Learning model storage unit]
The learning model storage unit 201 stores a learned learning model M in which the relationship between training product information on training products for training and training allocation information on the allocation of training products in a training store for training is learned. In this embodiment, since the learning model M is learned by the learning terminal 10, the learning model storage unit 201 stores the learning model M received by the server 20 from the learning terminal 10. When the server 20 also learns the learning model M, the learning model storage unit 201 stores the learning model M in which the server 20 has learned the training data. In this case, the server 20 has the same functions as the learning terminal 10. For example, the data storage unit 20 stores a training database DB1. The server 20 may include a learning unit 102.

[Estimated product information acquisition department]
The estimated product information acquisition unit 202 acquires estimated product information on estimated products handled in an estimated store for which shelf allocation is estimated. In this embodiment, the estimated product information is stored in the estimated database DB2, so the estimated product information acquisition unit 202 acquires the estimated product information from the estimated database DB2. For example, when a person in charge operates the person in charge terminal 30 to request a shelf allocation proposal from the shelf allocation proposal service, the estimated product information acquisition unit 202 acquires estimated product information associated with the store ID of the store for which shelf allocation is proposed from the estimated database DB2. The store for which shelf allocation is proposed is specified by the person in charge. Note that the estimated product information acquisition unit 202 may acquire the estimated product information from a database other than the estimated database DB2, the person in charge terminal 30, another terminal, or an information storage medium.

[Estimated shelf information acquisition unit]
The estimated shelf information acquisition unit 203 acquires estimated shelf information related to estimated shelves arranged in an estimated store. In this embodiment, since the estimated shelf information is stored in the estimation database DB2, the estimated shelf information acquisition unit 203 acquires the estimated shelf information from the estimation database DB2. For example, when a person in charge operates the person in charge terminal 30 to request a shelf allocation proposal to the shelf allocation proposal service, the estimated shelf information acquisition unit 203 acquires, from the estimation database DB2, estimated shelf information associated with the store ID of the store that is the target of the shelf allocation proposal. Note that the estimated shelf information acquisition unit 203 may acquire the estimated shelf information from a database other than the estimation database DB2, the person in charge terminal 30, another terminal, or an information storage medium.

[Estimated purchase information acquisition department]
The estimated purchase information acquisition unit 204 acquires estimated purchase information related to purchases at an estimated store. In this embodiment, the estimated purchase information is stored in the estimated database DB2, so the estimated purchase information acquisition unit 204 acquires the estimated purchase information from the estimated database DB2. For example, when a person in charge operates the person in charge terminal 30 to request a purchase discount proposal from the purchase discount proposal service, the estimated purchase information acquisition unit 204 acquires, from the estimated database DB2, estimated purchase information associated with the store ID of the store to which the purchase discount proposal is to be made. Note that the estimated purchase information acquisition unit 204 may acquire the estimated purchase information from a database other than the estimated database DB2, the person in charge terminal 30, another terminal, or an information storage medium.

[Proposal Department]
The suggestion unit 205 makes a suggestion regarding the shelf allocation of the estimated product in the estimated store based on the learning model M and the estimated product information. In this embodiment, since the learning model M has also learned training shelf information regarding training shelves arranged in the training store, the suggestion unit 205 makes a suggestion further based on the estimated shelf information. Furthermore, in this embodiment, since the learning model M has also learned training purchase information regarding purchases in the training store, the suggestion unit 205 makes a suggestion further based on the training purchase information.

For example, the proposal unit 205 inputs estimated data including estimated product information, estimated shelf information, and estimated purchase information to the learning model M. When the estimated data is input, the learning model M calculates an embedded representation of the estimated data based on parameters adjusted by learning. The learning model M outputs estimated shelf allocation information according to the embedded representation. The proposal unit 205 makes a proposal regarding shelf allocation to the person in charge based on the estimated shelf allocation information output from the learning model M. For example, the proposal unit 205 makes a proposal regarding shelf allocation by displaying a shelf allocation proposal screen SC for proposing shelf allocation on the person in charge terminal 30 based on the estimated shelf allocation information.

Figure 6 is a diagram showing an example of a shelf allocation proposal screen SC. In this embodiment, the learning model M has learned training promotion/demotion information regarding promotion and demotion regarding shelf allocation in the training store as training shelf allocation information. Therefore, the proposal unit 205 acquires estimated promotion/demotion information regarding promotion and demotion regarding shelf allocation in the estimated store based on the learning model and the estimated product information, and makes a proposal based on the estimated promotion/demotion information. For example, as shown in Figure 6, the proposal unit 205 displays a shelf allocation proposal screen SC showing the estimated promotion/demotion information on the staff terminal 30.

In this embodiment, the learning model M has learned training substitute product information regarding training substitute products that are substituted for the training product in the shelf allocation at the training store as training shelf allocation information. For example, the proposal unit 205 acquires estimated substitute product information that is substituted for the estimated product in the shelf allocation at the estimated store based on the learning model and the estimated product information, and makes a proposal based on the estimated substitute product information. For example, as shown in FIG. 6, the proposal unit 205 displays a shelf allocation proposal screen SC showing the estimated substitute product information on the staff terminal 30.

The suggestion unit 205 may make a suggestion by a method other than displaying the shelf allocation suggestion screen SC. For example, the suggestion unit 205 may make a suggestion by using a communication means such as email. The suggestion unit 205 may also make a suggestion based on information other than the training promotion/demotion information and the training substitute product information. For example, the suggestion unit 205 may make a suggestion based on estimated shelf allocation information indicating a list of estimated products to be placed on the estimated shelf. Furthermore, the suggestion unit 205 may make a suggestion for each space of the estimated shelf (for example, each space on the upper, middle, and lower levels) based on estimated shelf allocation information indicating a list of estimated products to be placed in that space. The suggestion unit 205 may make a suggestion based on estimated shelf allocation information indicating a list of estimated products that should not be placed on the estimated shelf.

[3-3. Functions realized on the person in charge terminal]
For example, the person in charge terminal 30 includes a data storage unit 300, an operation reception unit 301, and a display control unit 302. The data storage unit 300 is realized by the storage unit 32. The operation reception unit 301 and the display control unit 302 are realized by the control unit 31.

[Data storage unit]
The data storage unit 300 stores data necessary for the person in charge to use the shelf allocation proposal service. For example, the data storage unit 300 stores an application dedicated to the shelf allocation proposal service. When the person in charge uses the payment service from a browser instead of the application, the data storage unit 300 stores the browser.

[Operation reception section]
The operation reception unit 301 receives various operations from the user. For example, the operation reception unit 301 receives operations for an application or a browser dedicated to the shelf allocation proposal service. The operation reception unit 301 transmits data indicating the content of the user's operation to the server 20.

[Display control unit]
The display control unit 302 displays various screens on the display unit 35. For example, the display control unit 302 displays a shelf allocation proposal screen SC on the display unit 35. The display control unit 302 communicates with the server 20, receives data necessary for displaying the shelf allocation proposal screen SC, and displays the shelf allocation proposal screen SC on the display unit 35.

[4. Processing Executed by Shelf Planogram Proposal System]
Fig. 7 is a diagram showing an example of a process executed in the shelf allocation proposal system 1. The control units 11, 21, and 31 execute programs stored in the storage units 12, 22, and 32, respectively, to execute the process in Fig. 7.

As shown in FIG. 7, the learning terminal 10 learns the learning model M based on the training database DB1 (S1). The learning terminal 10 uploads the learned learning model M to the server 20 (S2). When the server 20 receives the learned learning model M from the learning terminal 10 (S3), it records the learned learning model M in the memory unit 22 (S4). The staff terminal 30 executes a login process between the server 20 and the staff based on the staff's operation, for the staff to log in to the shelf allocation proposal service (S5).

Based on the operation of the person in charge, the person in charge terminal 30 transmits estimated data including estimated product information, estimated shelf information, and estimated purchase information to the server 20 (S6). The server 20 receives the estimated data including estimated product information, estimated shelf information, and estimated purchase information from the person in charge terminal 30 (S7). The server 20 stores the estimated data in the estimation database DB2 (S8). Based on the operation of the person in charge, the person in charge terminal 30 requests the server 20 to propose a shelf allocation (S9). When the server 20 receives the request for a shelf allocation proposal from the person in charge terminal 30 (S10), it obtains the estimated data including estimated product information, estimated shelf information, and estimated purchase information from the estimation database DB2 (S11).

The server 20 inputs the estimated data acquired in S11 to the learning model M (S12). The server 20 acquires the estimated shelf allocation information output from the learning model M (S13). The server 20 generates display data for the shelf allocation proposal screen SC based on the estimated shelf allocation information and transmits it to the staff terminal 30 (S14). When the staff terminal 30 receives the display data for the shelf allocation proposal screen SC from the server 20 (S15), it displays the shelf allocation proposal screen SC on the display unit 35 (S16), and this process ends.

5. Summary of the embodiment
The shelf allocation proposal system 1 of the present embodiment stores a learned learning model M in which the relationship between the training product information and the training shelf allocation information has been learned. The shelf allocation proposal system 1 acquires estimated product information. The shelf allocation proposal system 1 makes a proposal regarding the shelf allocation of the estimated product in the estimated store based on the learning model M and the estimated product information. This allows the shelf allocation proposal system 1 to improve the accuracy of the shelf allocation proposal. For example, the shelf allocation proposal system 1 can propose shelf allocation based on the learning model M that can also handle unknown estimated product information, rather than proposing shelf allocation within a range of a predetermined shelf allocation pattern or relationship information as described in the prior art document. When a person in charge considers the shelf allocation of a new product, if the learning model M has learned training data of other products with similar characteristics to the new product, the learning model M can propose an appropriate shelf allocation of the new product based on parameters adjusted by the training data. Since the learning model M has learned training data of various training stores, the shelf allocation proposal system 1 can propose an appropriate shelf allocation that takes into account the tendencies of other stores other than the estimated store where the person in charge considers the shelf allocation.

In this embodiment, the learning model M also learns training shelf information. The shelf allocation proposal system 1 acquires estimated shelf information. The shelf allocation proposal system 1 makes a proposal further based on the estimated shelf information. This allows the shelf allocation proposal system 1 to propose an appropriate shelf allocation based not only on estimated product information of the estimated product that is the subject of shelf allocation, but also on estimated shelf information of the estimated shelf that is the subject of shelf allocation, thereby further improving the accuracy of shelf allocation proposals. For example, the shelf allocation proposal system 1 can propose an appropriate shelf allocation according to characteristics such as the size or capacity of the estimated shelf.

In this embodiment, the learning model M also learns training purchase information. The shelf allocation proposal system 1 acquires estimated purchase information. The shelf allocation proposal system 1 makes a proposal further based on the estimated purchase information. This allows the shelf allocation proposal system 1 to propose an appropriate shelf allocation based not only on estimated product information of the estimated product that is the subject of shelf allocation, but also on estimated purchase information of the estimated store that is the subject of shelf allocation, thereby further improving the accuracy of the shelf allocation proposal. For example, the shelf allocation proposal system 1 can propose an appropriate shelf allocation according to the characteristics of the customer base or best-selling products in the estimated store.

In addition, in this embodiment, the learning model M has learned training promotion/demotion information as training shelf allocation information. The shelf allocation proposal system 1 acquires estimated promotion/demotion information based on the learning model M and estimated product information. The shelf allocation proposal system 1 makes a proposal based on the estimated promotion/demotion information. This allows the shelf allocation proposal system 1 to propose promotion and demotion of estimated products. For example, since the person in charge can easily grasp which estimated products should be promoted and demoted based on the current shelf allocation, the shelf allocation proposal system 1 can reduce the workload of the person in charge.

In addition, in this embodiment, the learning model M has learned training substitute product information as training shelf allocation information. The shelf allocation proposal system 1 acquires estimated substitute product information based on the learning model M and the estimated product information. The shelf allocation proposal system 1 makes a proposal based on the estimated substitute product information. This allows the shelf allocation proposal system 1 to propose an estimated substitute product to replace the estimated product. For example, since the person in charge can easily grasp the estimated product to replace the estimated product placed on the estimated shelf in the current shelf allocation, the shelf allocation proposal system 1 can reduce the workload of the person in charge.

[6. Modifications]
The present disclosure is not limited to the above-described embodiments. The present disclosure can be modified as appropriate without departing from the spirit and scope of the present disclosure.

Figure 8 is a diagram showing an example of functions realized in the modified example. For example, the server 20 includes an other estimated purchase information acquisition unit 206, another estimated shelf information acquisition unit 207, and an expected customer information acquisition unit 208. Each of the other estimated purchase information acquisition unit 206, the other estimated shelf information acquisition unit 207, and the expected customer information acquisition unit 208 is realized by the control unit 21.

[6-1. Modification 1]
For example, the training purchase information may be information regarding purchases by training customers who visit a training store. A training customer who visits a training store is a customer who actually purchases training products at the training store. The training purchase information indicates information regarding purchases of at least one of training products placed on a training shelf for which the learning model M should output estimated shelf allocation information during learning, and training products placed on other training shelves in the training store. For example, the training purchase information indicates the quantity of these training products purchased, the total amount, the number of training customers, the characteristics of the training customers (e.g., gender or age group), or other characteristics. The training purchase information may indicate changes in these over time.

The learning model M of the first modified example also learns other training purchase information related to purchases by other training customers related to the training customer who have not visited the training store. In the first modified example, the input portion of the training data includes other training purchase information. The learning unit 102 makes the learning model M learn the training data including the other training purchase information in the input portion. Although the input portion of the training data is different from the embodiment, the learning method of the learning model M may be the same as the embodiment. The learning unit 102 records the learned learning model M that has learned the other training purchase information in the data storage unit 200. The learning unit 102 uploads the learned learning model M to the server 20.

The other training customers are customers whose characteristics are similar to those of the training customer. For example, the other training customers are customers whose gender, age group, tastes, behavior, or other characteristics are similar to those of the training customer. Also, for example, the other training customers may be customers whose attribute values are common to the training customer in at least some attributes including gender, age group, tastes, behavior, or other characteristics. The shelf allocation proposal system 1 may identify other training customers from information obtained in a physical store, or may identify other training customers from information obtained from a service other than a physical store. The other training purchase information is training purchase information of other training customers. The content of the training purchase information is as described in the embodiment. The other training purchase information differs from the training purchase information in that it indicates the purchasing characteristics of other training customers, but may be similar to the training purchase information in other respects. For example, the other training purchase information may indicate the gender, age group, tastes, behavior, or other characteristics of other training customers.

For example, the shelf allocation proposal system 1 may identify other training customers with similar characteristics to the training customer based on a service database in which usage status data indicating the usage status of users who use electronic commerce services, communication services, travel reservation services, electronic payment services, financial services, online flea market services, or other online services is accumulated. For example, the shelf allocation proposal system 1 may analyze various data covering a large number of items such as the attributes, purchasing trends, price preferences, and service usage trends of users who have a history of purchasing the target product, and extract users with similar characteristics from a user group with no purchase history as similar users. The training customer and other training customers may use these services, or may not particularly use these services. The shelf allocation proposal system 1 may identify what characteristics customers have that are similar to each other based on the service database.

For example, the shelf allocation proposal system 1 identifies users who have similar characteristics by clustering the usage data stored in the service database. The clustering method may be a known method. For example, the clustering method may be a method such as DB-SCAN or k-nearest neighbor method. For example, the shelf allocation proposal system may use other methods such as collaborative filtering to identify other training customers whose characteristics are similar to those of the training customer. When clustering is performed, belonging to the same cluster corresponds to having similar characteristics.

For example, the shelf allocation proposal system 1 identifies what characteristics customers have that are similar to each other based on the service database, and records a similar customer database indicating the identification results in the data storage unit 100. The learning unit 102 identifies other training customers that have similar characteristics to the training customer based on the similar customer database. The learning unit 102 trains the learning model M not only on training data generated based on the training customer, but also on training data generated based on other training customers. This allows the learning model M to learn more training data, thereby improving the accuracy of the learning model M.

For example, there may be a correlation between the characteristics of not only training customers who actually visited the training store, but also other training customers who have similar characteristics to the training customer, and the training allocation information that indicates the correct allocation for the training store. The learning model M learns the correlation between these by studying the training data. As a result, when the proposal unit 205 inputs estimated data including other estimated purchase information described below to the learning model M, the learning model M can output estimated promotion/demotion information and estimated substitute product information as estimated allocation information according to the estimated data, based on the correlation it has learned (i.e., by processing based on parameters adjusted by learning).

For example, information obtained from training customers who actually visited a training store may not be enough to accurately identify information such as the customer demographic that prefers that training store. For this reason, the shelf allocation proposal system 1 of variant example 1 can have the learning model M learn correlations that cannot be obtained from the training store alone by also using other training purchase information identified by a large amount of usage in services such as online services. For example, if the other training purchase information indicates a best-selling product in a service such as an online service, the learning model M can learn the correlation between that product and an appropriate shelf allocation. When estimated data including other estimated purchase information described below is input, the learning model M becomes able to output according to the correlations it has learned.

For example, the estimated purchase information may be information regarding purchases by estimated customers who visited an estimated store. An estimated customer who visited an estimated store is a customer who actually purchased an estimated product at the estimated store. The estimated purchase information indicates information regarding purchases of at least one of an estimated product placed on an estimated shelf for which the learning model M should output estimated shelf allocation information during learning, and an estimated product placed on another estimated shelf in the estimated store. For example, the estimated purchase information indicates the quantity of these estimated products purchased, the total amount, the number of estimated customers, the characteristics of the estimated customers (e.g., gender or age group), or other characteristics. The estimated purchase information may also indicate changes in these over time.

The shelf allocation proposal system 1 of the first modified example includes another estimated purchasing information acquisition unit 206. The other estimated purchasing information acquisition unit 206 acquires other estimated purchasing information related to purchases by other estimated customers who are related to the estimated customer and have not visited the estimated store. The other estimated customers are customers whose characteristics are similar to those of the estimated customer. For example, the other estimated customers are customers whose gender, age group, preferences, behavior, or other characteristics are similar to those of the estimated customer. The other estimated purchasing information acquisition unit 206 may identify other estimated customers whose characteristics are similar to those of the estimated customer in a manner similar to the method for identifying other training customers. This method is as described above.

For example, the data storage unit 200 may store the similar customer database described above. The other estimated purchase information acquisition unit 206 may identify other estimated customers with similar characteristics to the estimated customer based on the similar customer database, and acquire other estimated purchase information of the other estimated customers. The other estimated purchase information is estimated purchase information of the other estimated customers. The content of the estimated purchase information is as described in the embodiment. The other estimated purchase information differs from the estimated purchase information in that it indicates the purchasing characteristics of the other estimated customers, but may be similar to the estimated purchase information in other respects. In the first modified example, the other estimated purchase information is stored in the estimated database DB2. The other estimated purchase information acquisition unit 206 acquires the other estimated purchase information from the estimated database DB2. The other estimated purchase information acquisition unit 206 may acquire the other estimated purchase information from a database other than the estimated database DB2, the person in charge terminal 30, another terminal, or an information storage medium.

The other training customers may be customers whose embedded expressions based on gender, age group, preferences, or other characteristics are similar to those of the training customer. The embedded expressions of the other training customers satisfy predetermined conditions indices such as Euclidean distance and cosine similarity with the embedded expressions of the training customers. The shelf allocation proposal system 1 may acquire the embedded expressions by inputting attribute data including the gender, age group, preferences, or other characteristics of the target customers into a dedicated learning model, determine the similarity of the embedded expressions between customers, and identify the other training customers. Here, the dedicated learning model may be a pre-trained model including a user model such as UserBERT, or may be a pre-trained model including a graph neural network in which relationships between customers are learned by representation learning.

The shelf allocation proposal system 1 may dynamically adjust the similarity judgment conditions (similarity identification conditions) of the training customer corresponding to the training product according to the sales of the training product indicated by the training purchase information. For example, when the purchase quantity of the training product exceeds or falls below a predetermined value, the shelf allocation proposal system 1 may identify a customer whose attribute values in at least some of the above attributes are common to the training customer as the other training customer. Also, for example, when the purchase quantity of the training product exceeds or falls below a predetermined value, the shelf allocation proposal system 1 may adjust various parameters such as the number of neighboring objects so that other training customers similar to the training customer are increased, and perform the above clustering. Also, for example, when the purchase quantity of the training product exceeds or falls below a predetermined value, the shelf allocation proposal system 1 may adjust the threshold value of an index such as Euclidean distance or cosine similarity between the embedded representations of the customers so that other training customers similar to the training customer are increased, and perform a similarity judgment of other training customers. In addition, when an estimated product is given in advance, the shelf allocation proposal system 1 may adjust the similarity judgment conditions described above according to a training product that is the same as the estimated product, generate or expand the training data, and perform learning or additional learning of the learning model M.

The suggestion unit 205 of the first modification makes a suggestion further based on other estimated purchase information. The suggestion unit 205 inputs estimated data including other estimated purchase information to the learning model M. When the example of the embodiment and the first modification are combined, the suggestion unit 205 inputs estimated data including estimated product information, estimated shelf information, estimated purchase information, and other estimated purchase information. Although the estimated data input to the learning model M is different from the embodiment, the processing of the learning model M may be similar to the embodiment. The learning model M calculates an embedded expression of the estimated data, and outputs estimated shelf allocation information according to the embedded expression. The suggestion unit 205 makes a proposal regarding shelf allocation to the person in charge based on the estimated shelf allocation information output from the learning model M taking into account the other estimated purchase information.

The learning model M of the first modification also learns other training purchase information related to purchases by other training customers who have not visited the training store. The shelf allocation proposal system 1 acquires other estimated purchase information related to purchases by other estimated customers who have not visited the estimated store. The shelf allocation proposal system 1 makes a proposal based on the other estimated purchase information. As a result, the shelf allocation proposal system 1 can propose an appropriate shelf allocation based not only on the estimated customers who visited the estimated store that is the target of shelf allocation, but also on the other estimated purchase information of other estimated customers, so that the accuracy of the shelf allocation proposal can be further improved. For example, the shelf allocation proposal system 1 can propose an appropriate shelf allocation according to the characteristics of other estimated customers who have similar characteristics to the estimated customer who visited the estimated store. The shelf allocation proposal system 1 can propose an appropriate shelf allocation that also takes into account information obtained from services such as online services. For example, when a trend of other training customers similar to the training customer who visits the training store is indicated in the other training purchase information and the trend cannot be identified only by the training store, the shelf allocation proposal system 1 can propose an appropriate shelf allocation according to the trend.

[6-2. Modification 2]
For example, in the first modified example, the learning model M may learn other training purchase information of other training customers who are in the same trade area as the training customer. The trade area is an area targeted in a business. Being in the same trade area as the training customer means being in the same trade area as the training store, or being in a trade area within the range of the training customer's usual activities. For example, the trade area may be specified based on the communication status of mobile communication. The service provider of the shelf allocation proposal service may determine the trade area, or a trade area defined by an external service may be used. The trade area definition data that defines the trade area is stored in the data storage unit 100 and the data storage unit 200. For example, the trade area definition data indicates areas of the same trade area among areas all over the country. The learning unit 102 may identify other training customers who are in the same trade area as the training customer who visited a certain training store based on the trade area definition data, and generate training data.

For example, the input portion of the training data in variant example 2 includes other training purchase information of other training customers who are in the same trade area as a training customer who visited a certain training store. The learning unit 102 trains the learning model M with the training data including the other training purchase information in the input portion. This variant example differs from variant example 1 in that the input portion of the training data does not include other training purchase information of other training customers who are in a different trade area from the training customer who visited a certain training store, but the method of training the learning model M with the other training purchase information may be the same as variant example 1.

The other estimated purchasing information acquisition unit 206 of the second modification acquires other estimated purchasing information of other estimated customers who are in the same trade area as the estimated customer. For example, the other estimated purchasing information acquisition unit 206 identifies other estimated customers who are in the same trade area as the estimated customer who visited the estimated store in which the estimated shelf to be proposed for shelf allocation is located, based on the trade area definition data. The other estimated purchasing information acquisition unit 206 refers to the estimated database DB2 and acquires other estimated purchasing information of the identified other estimated customers. This is different from the first modification in that other estimated purchasing information of other estimated customers who are in a trade area different from the estimated customer who visited the estimated store in which the estimated shelf to be proposed for shelf allocation is located is not acquired, but the content of the other estimated purchasing information itself may be the same as that of the first modification.

The suggestion unit 205 of the second modification makes a proposal based on other estimated purchase information of other estimated customers in the same trade area as the estimated customer. The suggestion unit 205 inputs estimated data including other estimated purchase information of other estimated customers in the same trade area as the estimated customer to the learning model M. This modification differs from the first modification in that the estimated data does not include other estimated purchase information of other estimated customers in a trade area different from that of the estimated customer who visited the estimated store in which the estimated shelf for which the shelf allocation is proposed is located, but the processing of the learning model M may be the same as that of the first modification. The suggestion unit 205 makes a proposal regarding shelf allocation to the person in charge based on the estimated shelf allocation information output from the learning model M, taking into consideration other estimated purchase information of other estimated customers in the same trade area.

The learning model M of the second modification example has learned other training purchase information of other training customers who are in the same commercial area as the training customer. The shelf allocation proposal system 1 acquires other estimated purchase information of other estimated customers who are in the same commercial area as the estimated customer. The shelf allocation proposal system 1 makes a proposal based on other estimated purchase information of other estimated customers who are in the same commercial area as the estimated customer. In this way, the shelf allocation proposal system 1 can propose an appropriate shelf allocation based on other estimated purchase information of other estimated customers who are in the same commercial area as the estimated customer who visited the estimated store that is the target of shelf allocation, thereby further improving the accuracy of the shelf allocation proposal. For example, the shelf allocation proposal system 1 can propose an appropriate shelf allocation according to the characteristics of other estimated customers who are more similar in characteristics to the estimated customer who visited the estimated store.

[6-3. Modification 3]
For example, as described somewhat in the embodiment, the training purchase information may be information about a training customer who purchased a training product on a training shelf arranged in a training store. The training purchase information of the modified example 3 indicates the characteristics of the training customer himself, not the characteristics of the training product purchased by the training customer. For example, the training purchase information indicates the gender, age group, preferences, behavior, or other characteristics of the training customer. The other characteristics may be characteristics used as demographic information. For example, the training purchase information may indicate the residential area, range of activities, occupation, or annual income of the training customer. In the modified example 3, it is assumed that such training purchase information is stored in the training database DB1. The learning unit 102 of the modified example 3 learns the learning model M based on such training purchase information. The learning method of the learning model M may be the same as that of the embodiment.

The estimated purchase information of the third modification is information about an estimated customer who purchased an estimated product on an estimated shelf arranged in an estimated store. The estimated purchase information of the third modification indicates the characteristics of the estimated customer himself, not the characteristics of the estimated product purchased by the estimated customer. For example, the estimated purchase information indicates the gender, age group, preferences, behavior, or other characteristics of the estimated customer. The other characteristics may be characteristics used as demographic information. For example, the estimated purchase information may indicate the residential area, range of activities, occupation, or annual income of the estimated customer. In the third modification, such estimated purchase information is stored in the estimated database DB2. The suggestion unit 205 of the third modification proposes a shelf allocation based on such estimated purchase information. The processing of the learning model M during estimation may be the same as in the embodiment.

In variant example 3, the training purchase information is information about a training customer who purchased a training product on a training shelf placed in a training store. The estimated purchase information is information about an estimated customer who purchased an estimated product on an estimated shelf placed in an estimated store. This allows the shelf allocation proposal system 1 to propose shelf allocations according to the estimated customers, thereby further improving the accuracy of shelf allocation proposals. For example, the shelf allocation proposal system 1 can propose appropriate shelf allocations according to the characteristics, such as the gender or age group, of the estimated customers who visited the estimated store.

[6-4. Modification 4]
For example, the learning model M may have learned the relationship between training product information and training shelf allocation information for each possible placement location where training products can be placed on a training shelf arranged in a training store. A possible placement location is an individual space on the training shelf. The training shelf is divided into a plurality of possible placement locations. For example, each tier of the training shelf, such as the top tier, middle tier, and bottom tier, is a possible placement location. Within each tier, horizontal positions such as the left side, center, and right side are also possible placement locations. The number of possible placement locations that a training shelf has may be any number.

The training data of the fourth modification includes a pair of an input part and an output part for each possible placement location on a training shelf. For example, if a training shelf is divided vertically into three parts, the top, middle, and bottom, and horizontally into three parts, the left, center, and right, the training shelf includes nine possible placement locations. The training data of the training shelf includes a pair of an input part and an output part for each of the nine possible placement locations. Thus, one training data includes nine pairs. The learning unit 102 of the fourth modification causes such training data to be learned by the learning model M. In the fourth modification, a boosting model (decision tree model) may be used as the learned learning model M. The learning model M may process a classification task that solves the classification of the installation location for each estimated product.

For example, the first input part of the training data is training product information, etc. of a training product that is a candidate for placement in the possible placement location on the left side of the upper row. The output part corresponding to the input part is training allocation information that indicates an appropriate shelf allocation for the possible placement location on the left side of the upper row. The second input part of the training data is training product information, etc. of a training product that is a candidate for placement in the possible placement location in the center of the upper row. The output part corresponding to the input part is training allocation information that indicates an appropriate shelf allocation for the possible placement location in the center of the upper row. Similarly, the third to ninth input and output parts of the training data are training product information, etc. of a training product that is a candidate for placement in the possible placement locations on the right side of the upper row, the left side of the middle row, the center of the middle row, the right side of the middle row, the left side of the lower row, the center of the lower row, and the right side of the lower row, and training allocation information that indicates an appropriate shelf allocation for the possible placement locations.

The shelf allocation proposal system 1 of the modified example 4 includes a placement information acquisition unit. The placement information acquisition unit acquires placement location information regarding possible locations where the estimated product can be placed on the estimated shelf arranged in the estimated store. The meaning of the possible placement locations of the estimated shelf is the same as that of the possible placement locations of the training shelf. Here, as with the possible placement locations of the training shelf, it is assumed that a certain estimated shelf is divided into three vertically, the upper, middle, and lower levels, and three horizontally, the left, center, and right sides. The estimated shelf has nine possible placement locations. In the modified example 4, a case will be described in which the possible placement locations of the estimated shelf and the possible placement locations of the training shelf are the same, but these possible placement locations may differ to some extent. In the modified example 4, it is assumed that such possible placement information is stored in the estimated database DB2. The possible placement information acquisition unit acquires the possible placement information from the estimated database DB2. The placement information acquisition unit may acquire placement information from a database other than the estimated database DB2, the person in charge terminal 30, another terminal, or an information storage medium.

The suggestion unit 205 of the fourth modification makes a proposal based further on the possible placement location information. The suggestion unit 205 inputs estimated data including possible placement location information to the learning model M. When the example of the embodiment and the fourth modification are combined, the suggestion unit 205 inputs estimated data including estimated product information, estimated shelf information, estimated purchase information, and possible placement location information. Although the estimated data input to the learning model M is different from the embodiment, the processing of the learning model M may be similar to the embodiment. The learning model M calculates an embedded expression of the estimated data and outputs estimated shelf allocation information according to the embedded expression. The estimated shelf allocation information indicates an appropriate shelf allocation for each possible placement location of the estimated shelf. The suggestion unit 205 makes a proposal regarding shelf allocation to the person in charge based on the estimated shelf allocation information output from the learning model M taking into account the possible placement location information.

In the learning model M of variant example 4, the relationship between the training product information and the training shelf allocation information is learned for each possible placement location. The shelf allocation proposal system 1 acquires the possible placement location information of the estimated store. The shelf allocation proposal system 1 makes a proposal further based on the possible placement location information. This allows the shelf allocation proposal system 1 to make shelf allocation proposals for each possible placement location, thereby further improving the accuracy of shelf allocation proposals. For example, the person in charge can grasp the optimal shelf allocation for each possible placement location, so the shelf allocation proposal system 1 can increase the convenience of the person in charge.

[6-5. Modification 5]
For example, the learning model M may also learn other training shelf information related to other training shelves around the training shelf arranged in the training store. The other training shelves are other training shelves arranged in the same training store. Surrounding means that the training shelf that is the subject of shelf allocation in the training data is close in distance. For example, other training shelves next to the training shelf that is the subject of shelf allocation in the training data, in the same sales area, or on the same floor are other training shelves in the surroundings. Surrounding training shelf data indicating other training shelves in the surroundings is stored in the data storage unit 100 and the data storage unit 200. For example, the surrounding training shelf data indicates whether training shelves in a certain training store are in the surroundings of each other. The learning unit 102 may identify other training shelves around a certain training shelf based on the surrounding training shelf data and generate training data.

For example, the input portion of the training data in variant 5 includes other training shelf information of other training shelves around a certain training shelf. The other training shelf information differs from the training shelf information in that it indicates the characteristics of the other training shelves, but may be similar to the training shelf information in other respects. For example, the other training shelf information may indicate the size, capacity, number of tiers, type, design, material, location in the store, functions such as refrigerated/non-refrigerated, or other characteristics of the other training shelves. The other training shelf information may indicate training products arranged on other training shelves. For example, the other training shelf information may indicate product identification information such as product name, content, type, price, size, design, taste, texture, JAN code, or other characteristics of the training products arranged on other training shelves. The learning unit 102 causes the learning model M to learn the training data including the other training shelf information in the input portion. This differs from the embodiment in that the other training shelf information is not included in the input portion of the training data, but the method of causing the learning model M to learn the training data may be similar to the embodiment.

For example, a shelf allocation appropriate for a certain training shelf may affect the shelf allocation of other training shelves. It may be better for a certain training shelf not to have training products or similar products placed on other surrounding training shelves. Conversely, depending on the specific type of training product, it may be better for a certain training shelf to have training products or similar products placed on other surrounding training shelves. For this reason, a correlation may exist between other training shelf information of other training shelves around a certain training shelf and training shelf allocation information that shows a shelf allocation appropriate for a certain training shelf. By learning such correlations, the learning model M of variant example 5 becomes able to output, at the time of estimation, according to other estimated shelves around a certain estimated shelf.

The shelf allocation proposal system 1 of the modified example 5 includes an other estimated shelf information acquisition unit 207. The other estimated shelf information acquisition unit 207 acquires other estimated shelf information related to other estimated shelves around the estimated shelf arranged in the estimated store. The other estimated shelves are other estimated shelves arranged in the same estimated store. The meaning of "around" may be the same as that of the explanation of other training shelves. For example, other estimated shelves next to the estimated shelf that is the subject of shelf allocation in the estimated data, in the same sales area, or on the same floor are other estimated shelves in the surroundings. The other estimated shelf information acquisition unit 207 may identify other estimated shelves around a certain estimated shelf based on the surrounding training shelf data and generate estimated data. In the modified example 5, it is assumed that other estimated shelf information is stored in the estimated database DB2. The other estimated shelf information acquisition unit 207 acquires other estimated shelf information from the estimated database DB2. The other estimated shelf information acquisition unit 207 may acquire other estimated shelf information from a database other than the estimated database DB2, the person in charge terminal 30, another terminal, or an information storage medium.

The suggestion unit 205 of the fifth modification makes a suggestion further based on other estimated shelf information. The suggestion unit 205 inputs estimated data including other estimated shelf information to the learning model M. When the example of the embodiment and the fifth modification are combined, the suggestion unit 205 inputs estimated data including estimated product information, estimated shelf information, estimated purchase information, and other estimated shelf information. Although the estimated data input to the learning model M is different from the embodiment, the processing of the learning model M may be similar to the embodiment. The learning model M calculates an embedded expression of the estimated data, and outputs estimated shelf information according to the embedded expression. The suggestion unit 205 makes a suggestion regarding shelf allocation to the person in charge based on the estimated shelf allocation information output from the learning model M taking into account the other estimated shelf information.

The learning model M of variant example 5 also learns other training shelf information related to other training shelves around the training shelf placed in the training store. The shelf allocation proposal system 1 acquires other estimated shelf information. The shelf allocation proposal system 1 makes a proposal further based on the other estimated shelf information. This allows the shelf allocation proposal system 1 to propose a shelf allocation that also takes into account the characteristics of other estimated shelves around the estimated shelf, thereby further improving the accuracy of the shelf allocation proposal. For example, the shelf allocation proposal system 1 can propose a shelf allocation that takes into comprehensive consideration the relationship between the estimated shelf and other estimated shelves around it.

[6-6. Modification 6]
For example, the learning model M may also learn training customer information related to training customers who visited the training store. The training customer information may be the same as the training purchase information described in the embodiment and the first modified example. The training customer information may be information related to training customers who did not make a purchase at the training store (training customers who simply visited the training store and did not buy anything). The training customer information indicates the gender, age group, preferences, behavior, or other characteristics of these training customers. In the sixth modified example, the training customer information is stored in the training database DB1. For example, the input portion of the training data includes the training customer information. The learning unit 102 makes the learning model M learn the training data including the training customer information in the input portion. Although the input portion of the training data is different from the embodiment, the learning method of the learning model M may be the same as the embodiment. The learning unit 102 records the learned learning model M that has learned the training customer information in the data storage unit 200. The learning unit 102 uploads the learned learning model M to the server 20.

For example, there may be a correlation between the customer demographics of training customers who visited a training store and the appropriate shelf allocation for the training shelves placed in that training store. The appropriate shelf allocation for training shelves in a training store with many young female customers may differ from the appropriate shelf allocation for training shelves in a training store with many male office workers. By learning these correlations, the learning model M becomes able to output according to the customer demographics of the estimated store at the time of estimation. Although information on estimated customers who actually visited the estimated store may be used at the time of estimation, in variant example 6, information on expected customers who are expected to attract customers to the estimated store is used.

The shelf allocation proposal system 1 of the sixth modified example includes an expected customer information acquisition unit 208. The expected customer information acquisition unit 208 acquires expected customer information on expected customers that the estimated store is expected to attract. The expected customers are customers that the estimated store is expected to attract in the future. The expected customers may be designated by a store clerk or by a person in charge of shelf allocation. The expected customer information is information indicating the characteristics of the expected customers. For example, the expected customer information indicates the gender, age group, preferences, behavior, or other characteristics of the expected customers. In the sixth modified example, the expected customer information is stored in the estimated database DB2. The expected customer information acquisition unit 208 acquires the expected customer information from the estimated database DB2. The expected customer information acquisition unit 208 may acquire the expected customer information from a database other than the estimated database DB2, the person in charge terminal 30, another terminal, or an information storage medium.

The proposal unit 205 of the sixth modification makes a proposal further based on the expected customer information. The proposal unit 205 inputs estimated data including expected customer information to the learning model M. When the example of the embodiment and the sixth modification are combined, the proposal unit 205 inputs estimated data including estimated product information, estimated shelf information, estimated purchase information, and expected customer information. Although the estimated data input to the learning model M is different from the embodiment, the processing of the learning model M may be similar to the embodiment. The learning model M calculates an embedded expression of the estimated data, and outputs estimated shelf allocation information according to the embedded expression. The proposal unit 205 makes a proposal regarding shelf allocation to the person in charge based on the estimated shelf allocation information output from the learning model M taking into account the expected customer information.

The learning model M of variant example 6 also learns training customer information regarding training customers who visited the training store. The shelf allocation proposal system 1 acquires expected customer information. The shelf allocation proposal system 1 makes a proposal further based on the expected customer information. This allows the shelf allocation proposal system 1 to propose an appropriate shelf allocation according to expected customers expected at the estimated store that is the subject of shelf allocation, thereby further improving the accuracy of the shelf allocation proposal. For example, the shelf allocation proposal system 1 can propose a shelf allocation that will increase the future customer attraction effect of the estimated store.

[6-7. Other Modifications]
For example, the above modifications 1 to 6 may be combined.

For example, in the embodiment, an example has been given of a case where estimated shelf information and estimated purchase information are included in the estimated data, but a separate learning model M may be prepared for at least one of the estimated shelf information and the estimated purchase information. In this case, the proposal unit 205 may propose a shelf allocation based on a learning model M corresponding to the estimated shelf information out of the multiple learning models M. The proposal unit 205 may propose a shelf allocation based on a learning model M corresponding to the estimated purchase information out of the multiple learning models M.

For example, the functions described as being realized by the learning terminal 10 may be realized by the server 20, the staff member terminal 30, or another computer. The processes described as being realized by the learning terminal 10 may be shared among multiple computers. The processes described as being realized by the server 20 may be realized by the learning terminal 10, the staff member terminal 30, or another computer. The main functions of the shelf allocation proposal system 1 may be shared among multiple computers.

[7. Notes]
For example, the shelf layout proposal system can be configured as follows.
(1)
a learning model storage unit that stores a learned learning model in which a relationship between training product information on a training product for training and training shelf allocation information on the shelf allocation of the training product in a training store for training is learned;
an estimated product information acquisition unit that acquires estimated product information regarding estimated products handled in an estimated store whose shelf allocation is estimated;
A proposal unit that makes a proposal regarding a shelf allocation of the estimated product in the estimated store based on the learning model and the estimated product information;
A shelf layout proposal system that includes
(2)
The learning model also learns training shelf information regarding the training shelves arranged in the training store,
The shelf allocation proposal system further includes an estimated shelf information acquisition unit that acquires estimated shelf information regarding estimated shelves arranged in the estimated store,
The suggestion unit makes the suggestion further based on the estimated shelf information.
A shelf allocation proposal system as described in (1).
(3)
The learning model also learns training purchase information regarding purchases at the training store,
The shelf allocation proposal system further includes an estimated purchase information acquisition unit that acquires estimated purchase information regarding purchases in the estimated store,
The suggestion unit makes the suggestion further based on the estimated purchase information.
A shelf allocation proposal system according to (1) or (2).
(4)
The training purchase information is information regarding purchases made by training customers who visited the training store,
The estimated purchase information is information regarding purchases made by an estimated customer who visited the estimated store,
The learning model also learns other training purchase information related to purchases by other training customers related to the training customer who have not visited the training store,
The shelf allocation proposal system further includes a different estimated purchase information acquisition unit that acquires different estimated purchase information regarding purchases by different estimated customers related to the different estimated customer who have not visited the different estimated store,
The suggestion unit makes the suggestion further based on the other estimated purchase information.
A shelf allocation proposal system as described in (3).
(5)
The learning model learns the other training purchase information of the other training customers who are in the same trade area as the training customer,
The other estimated purchase information acquisition unit acquires the other estimated purchase information of the other estimated customers who are in the same trade area as the estimated customer,
the suggestion unit makes the suggestion based on the other estimated purchase information of the other estimated customers who are in the same trade area as the estimated customer.
A shelf allocation proposal system as described in (4).
(6)
The training purchase information is information about a training customer who purchased the training product on the training shelf arranged in the training store,
The estimated purchase information is information regarding an estimated customer who purchased the estimated product on the estimated shelf arranged in the estimated store.
A shelf allocation proposal system according to any one of (3) to (5).
(7)
The learning model learns a relationship between the training product information and the training shelf allocation information for each possible location where the training product can be arranged on the training shelf arranged in the training store,
The shelf allocation proposal system further includes a possible placement information acquisition unit that acquires possible placement location information regarding a possible placement location where the estimated product can be placed on the estimated shelf arranged in the estimated store,
The suggestion unit makes the suggestion further based on the possible placement location information.
A shelf allocation proposal system according to any one of (1) to (6).
(8)
The learning model also learns other training shelf information regarding other training shelves around the training shelf arranged in the training store,
The shelf allocation proposal system further includes an other estimated shelf information acquisition unit that acquires other estimated shelf information related to other estimated shelves around the estimated shelf arranged in the estimated store,
The suggestion unit makes the suggestion further based on the other estimated shelf information.
A shelf allocation proposal system according to any one of (1) to (7).
(9)
The learning model also learns training customer information regarding training customers who visit the training store,
The shelf allocation proposal system further includes an expected customer information acquisition unit that acquires expected customer information regarding expected customers that the estimated store is expected to attract,
The proposal unit makes the proposal further based on the expected customer information.
A shelf allocation proposal system according to any one of (1) to (8).
(10)
In the learning model, training promotion/demotion information regarding promotion and demotion regarding shelf allocation in the training store is learned as the training shelf allocation information,
The suggestion unit acquires estimated promotion/demotion information regarding promotion and demotion of shelf allocation in the estimated store based on the learning model and the estimated product information, and makes the suggestion based on the estimated promotion/demotion information.
A shelf allocation proposal system according to any one of (1) to (9).
(11)
In the learning model, as the training planogram information, training substitute product information regarding a training substitute product that is substituted for the training product in the planogram in the training store is learned,
The suggestion unit acquires information on an estimated substitute product to be substituted for the estimated product in a shelf allocation in the estimated store based on the learning model and the estimated product information, and makes the suggestion based on the estimated substitute product information.
A shelf allocation proposal system according to any one of (1) to (10).

1 Shelf allocation proposal system, 10 Learning terminal, 11, 21, 31 Control unit, 12, 22, 32 Memory unit, 13, 23, 33 Communication unit, 14, 34 Operation unit, 15, 35 Display unit, 20 Server, 30 Personnel terminal, 100, 200, 300 Data memory unit, 101, 201 Learning model memory unit, 102 Learning unit, 202 Estimated product information acquisition unit, 203 Estimated shelf information acquisition unit, 204 Estimated purchase information acquisition unit, 205 Proposal unit, 206 Other estimated purchase information acquisition unit, 207 Other estimated shelf information acquisition unit, 208 Expected customer information acquisition unit, 301 Operation reception unit, 302 Display control unit, DB1 Training database, DB2 Estimation database, N Network, M Learning model, SC Shelf allocation proposal screen.

Claims (11)

a learning model storage unit that stores a learned learning model in which a relationship between training product information on training products for training and training shelf allocation information on shelf allocation of the training products in a training store for training is learned, the learning model being learned to include training purchase information on purchases by training customers who visit the training store and other training purchase information on purchases by other training customers related to the training customers who do not visit the training store;
an estimated product information acquisition unit that acquires estimated product information on estimated products handled in an estimated store whose shelf allocation is estimated, the estimated product information being related to purchases by estimated customers who visit the estimated store;
an estimated purchase information acquisition unit that acquires estimated purchase information regarding purchases at the estimated store;
a second estimated purchase information acquisition unit that acquires second estimated purchase information related to purchases by other estimated customers related to the second estimated customer who have not visited the second estimated store;
a proposal unit that makes a proposal regarding a shelf allocation of the estimated product in the estimated store based on the learning model, the estimated product information, the estimated purchase information, and the other estimated purchase information;
A shelf layout proposal system that includes The learning model also learns training shelf information regarding the training shelves arranged in the training store,
The shelf allocation proposal system further includes an estimated shelf information acquisition unit that acquires estimated shelf information regarding estimated shelves arranged in the estimated store,
The suggestion unit makes the suggestion further based on the estimated shelf information.
The shelf layout proposal system according to claim 1. The learning model learns the other training purchase information of the other training customers who are in the same trade area as the training customer,
The other estimated purchase information acquisition unit acquires the other estimated purchase information of the other estimated customers who are in the same trade area as the estimated customer,
the suggestion unit makes the suggestion based on the other estimated purchase information of the other estimated customers who are in the same trade area as the estimated customer.
The shelf layout proposal system according to claim 1 or 2. The training purchase information is information about a training customer who purchased the training product on the training shelf arranged in the training store,
The estimated purchase information is information regarding an estimated customer who purchased the estimated product on the estimated shelf arranged in the estimated store.
The shelf layout proposal system according to claim 1 or 2. a learning model storage unit that stores a learned learning model in which a relationship between training product information on training products for training and training shelf allocation information on shelf allocation of the training products in a training store for training is learned, the learning model having learned the relationship between the training product information and the training shelf allocation information for each possible placement location where the training products can be placed on a training shelf arranged in the training store;
an estimated product information acquisition unit that acquires estimated product information regarding estimated products handled in an estimated store whose shelf allocation is estimated;
an arrangement possible location information acquisition unit that acquires arrangement possible location information regarding a possible location where the estimated product can be arranged on the estimated shelf arranged in the estimated store;
a proposal unit that makes a proposal regarding a shelf allocation of the estimated product in the estimated store based on the learning model, the estimated product information, and the possible placement location information;
A shelf layout proposal system that includes a learning model storage unit that stores a learned learning model in which a relationship between training product information on training products for training and training shelf allocation information on shelf allocation of the training products in a training store for training is learned, the learning model also learning other training shelf information on other training shelves around a training shelf arranged in the training store;
an estimated product information acquisition unit that acquires estimated product information regarding estimated products handled in an estimated store whose shelf allocation is estimated;
an other estimated shelf information acquisition unit that acquires other estimated shelf information related to other estimated shelves around the estimated shelf arranged in the estimated store;
A proposal unit that makes a proposal regarding the shelf allocation of the estimated product in the estimated store based on the learning model, the estimated product information, and the other estimated shelf information;
A shelf layout proposal system that includes a learning model storage unit that stores a learned learning model in which a relationship between training product information on training products for training and training shelf allocation information on shelf allocation of the training products in a training store for training is learned, the learning model also learning training customer information on training customers who visit the training store;
an estimated product information acquisition unit that acquires estimated product information regarding estimated products handled in an estimated store whose shelf allocation is estimated;
an expected customer information acquisition unit that acquires expected customer information regarding expected customers that the estimated store is expected to attract;
A proposal unit that makes a proposal regarding a shelf allocation of the estimated product in the estimated store based on the learning model, the estimated product information, and the expected customer information;
A shelf layout proposal system that includes a learning model storage unit that stores a learned learning model in which a relationship between training product information on training products for training and training allocation information on the allocation of the training products in a training store for training is learned, the learning model being learned with training promotion/demotion information on promotion and demotion regarding the allocation of the training products in the training store as the training allocation information;
an estimated product information acquisition unit that acquires estimated product information regarding estimated products handled in an estimated store whose shelf allocation is estimated;
a proposal unit that acquires estimated promotion/demotion information regarding promotion and demotion of shelf allocation in the estimated store based on the learning model and the estimated product information, and makes a proposal regarding shelf allocation of the estimated product in the estimated store based on the estimated promotion/demotion information;
A shelf layout proposal system that includes a learning model storage unit that stores a learned learning model in which a relationship between training product information on a training product for training and training allocation information on the allocation of the training product in a training store for training is learned, the learning model being learned with training substitute product information on a training substitute product that is substituted for the training product in the allocation of the training store as the training allocation information;
an estimated product information acquisition unit that acquires estimated product information regarding estimated products handled in an estimated store whose shelf allocation is estimated;
a proposal unit that acquires information on an estimated substitute product that is substituted for the estimated product in a shelf allocation in the estimated store based on the learning model and the estimated product information, and makes a proposal regarding the shelf allocation of the estimated product in the estimated store based on the estimated substitute product information;
A shelf layout proposal system that includes an estimated product information acquisition step of acquiring estimated product information on estimated products handled in an estimated store whose shelf allocation is estimated, the estimated product information being related to purchases by estimated customers who visit the estimated store;
an estimated purchase information acquisition step of acquiring estimated purchase information regarding purchases at the estimated store;
a step of acquiring other estimated purchase information regarding purchases by other estimated customers related to the estimated customer who have not visited the estimated store;
a training product information relating to a training product for training and training shelf allocation information relating to the shelf allocation of the training product in a training store for training, the training product information being a learned learning model in which a relationship has been learned, the training purchase information relating to a purchase by a training customer who visited the training store and other training purchase information relating to a purchase by another training customer related to the training customer who has not visited the training store, the training product information being a proposed shelf allocation of the estimated product in the estimated store based on the estimated product information, the estimated purchase information, and the other estimated purchase information;
A shelf allocation proposal method in which the above is executed by a computer . an estimated product information acquisition unit that acquires estimated product information on estimated products handled in an estimated store whose shelf allocation is estimated, the estimated product information being related to purchases by estimated customers who visit the estimated store;
an estimated purchase information acquisition unit that acquires estimated purchase information regarding purchases at the estimated store;
a second estimated purchase information acquisition unit that acquires second estimated purchase information related to purchases by other estimated customers related to the second estimated customer who have not visited the second estimated store;
a trained learning model in which a relationship between training product information on training products for training and training shelf allocation information on the shelf allocation of the training products in a training store for training is learned, the training purchase information on purchases by training customers who visit the training store and other training purchase information on purchases by other training customers related to the training customer but who do not visit the training store are learned, a proposal unit that makes a proposal regarding the shelf allocation of the estimated product in the estimated store based on the training model, the estimated product information, the estimated purchase information, and the other estimated purchase information;
A program that makes a computer function as a

Images