JP2023068360A - PARTS TYPE PREDICTION DEVICE AND METHOD - Google Patents

Abstract

To provide a component type prediction device and method for predicting a type with high accuracy by selecting a specification having a high correlation to the type from a large amount of specification information after selecting a long term delivery component having a risk of a missing component and an excessive stock.SOLUTION: A component type prediction device includes a storage part for storing receiving/shipping/stock information recording the number of past receiving/shipping/stock results of components constituting a product, and past item information recording specifications and types of past items, an object component selection part for selecting a component of a prediction object from the receiving/shipping/stock information, a specification grouping part for dividing specifications to be used into a plurality of groups from the past item information, a correlation degree measurement part for extracting a specification having a high use frequency from specification group information to calculate a correlation degree between the specification and a type, a specification selection part for combining specifications having a high correlation degree for each group from the correlation degree information and prediction object item information to select a specification that becomes necessary for prediction, and a type prediction part for predicting a type of the object from specification selection result information.SELECTED DRAWING: Figure 1

Description

TECHNICAL FIELD The present invention relates to a part model prediction device and method for managing product parts procurement.

For example, in the business of elevators and escalators as a product, production starts after receiving an order according to customer requirements, but in order to meet delivery deadlines, it is particularly necessary to prepare parts with long lead times in advance. However, since the model of parts cannot be decided until the design of the elevator is completed, it is difficult to prepare and causes shortages and excess inventory.

On the other hand, some specifications, such as elevator car size and speed, are determined before design details are finalized. Therefore, there is a technique for predicting the type of parts with a long lead time from such specification information and stocking them in advance. Patent Literature 1 states, “For items of specifications for which specification values have been input, the required quantity calculation unit 11 calculates the required quantities of parts based on the specification values while referring to the configuration information. item, the required amount of parts is calculated based on the predicted value while referring to the configuration information."

JP 2017-528326 A

In the technique of Patent Literature 1, it is necessary to manually select the parts for which the required amount is to be calculated. For this reason, if it is desired to improve the stocking of parts with the risk of stockouts or excess inventory using the technique of Patent Document 1, it is necessary to manually determine the risk of each part.

However, since elevators have thousands of different types of parts, they cannot be determined manually. In addition, the required amount of parts is calculated based on the known relationship between the specification values and the parts configuration of the product.

Therefore, it is necessary to investigate the relationship between the specification values and the parts configuration of the product in advance, but if there are many target parts, the preliminary investigation takes a lot of time. In addition, when the specification values are undetermined, the model is predicted using the result of the specification value prediction, but the prediction accuracy of the model depends on the prediction accuracy of the specification value, so there is a risk that the prediction accuracy will decrease. .

The present invention has been made in view of the above circumstances. It is an object of the present invention to provide a part type prediction device and method capable of accurately predicting the type.

Based on the above, in the present invention, "receiving/shipping/inventory information that records the number of past warehousing/shipping/inventory records of the parts that make up the product, and past item information that records the specifications and models of past items are stored. a storage unit that selects parts to be predicted from incoming/outgoing/inventory information; a specification grouping unit that classifies specifications into multiple groups based on past project information; , extracts specifications with high frequency of use, and calculates the degree of correlation between the specifications and the model, the correlation degree information, and the prediction target project information, combining specifications with a high degree of correlation for each group, necessary for prediction A parts model prediction device characterized by comprising a specification selection unit for selecting a specification to become and a model prediction unit for predicting a target model from specification selection result information.

In addition, in the present invention, "when reading the order information and predicting the type of an uncompleted item based on the specification information and the past delivery information, the past item is read and the type is predicted from among a plurality of specifications. Select specifications that are candidates to be used in the process, select specifications to be used for prediction from among the narrowed down candidates, build a prediction model, and use the specification information to determine the part type that has not yet been completed in the latest order. Part type prediction method characterized by predicting."

Advantageous Effects of Invention According to the present invention, it becomes easy to identify long-delivery parts that are at risk of being out of stock or overstocked, among elevator parts. Furthermore, for those parts, by appropriately selecting specifications that are correlated with the model from a huge number of specifications and predicting the model with high accuracy, it is possible to prepare parts without the risk of shortages or excess inventory.

Problems, configurations, and effects other than those described above will be clarified by the following description of the embodiments.

The figure which shows the functional block structural example of the component model prediction apparatus in a present Example. The figure which shows the flow of the process which the component model prediction apparatus 1 in a present Example performs. The figure which shows the data structure example of the warehousing/shipping/inventory information D1. The figure which shows the data structure example of the prediction component information D2. The figure which shows the data structure example of the past case information D3. The figure which shows the data structure example of the similar specification information D4. The figure which shows the data structure example of the unified specification information D5. The figure which shows the data structure example of the new past case information D6. The figure which shows the data structure example of the specification group information D7. The figure which shows the data structure example of the correlation information D8. The figure which shows the data structure example of the score information D9. The figure which shows the data structure example of the specification selection result information D10. The figure which shows the data structure example of the prediction object case information D11. The figure which shows the data structure example of the model prediction information D12. The figure which shows an example of an output screen.

Hereinafter, embodiments of the present invention will be described with reference to the drawings.

In this specification, a project means an order from a customer, and the registered information is composed of the customer name, delivery destination, and the like. However, it is not limited to this.

In the following description, an elevator is exemplified as a product, but other products can be widely applied as long as they are made-to-order products and the number of parts tends to be high-variety and low-volume. The effects of the present invention are remarkably exhibited.

An embodiment of a part type prediction device according to the present invention will be described below with reference to the drawings.

FIG. 1 is a diagram showing an example of functional blocks of a part type prediction device 1 according to an embodiment of the present invention. The part model prediction apparatus 1 of this embodiment is connected to a user terminal 50 used by a user and an external database DB2 storing data via a network Nw.

The user terminal 50 is an information processing device such as a PC (Personal Computer). A user issues a processing execution instruction to the component type prediction apparatus 1 configured by a computer through the user terminal 50 . The user terminal 50 also has a function of displaying information output by the part type prediction apparatus 1 to the user via the output unit 40 . The external database DB2 is, for example, a system such as ERP (Enterprise Resources Planning), a database storing data according to the system, or a storage device.

The network Nw communicably connects the user terminal 50, the database DB2, and the part model prediction apparatus 1. FIG. The network Nw is, for example, a LAN (Local Area Network), a WAN (Wide Area Network), a VPN (Virtual Private Network), or a communication network partially or wholly using a general public line such as the Internet.

The part model prediction device 1 is an information processing device such as a PC or a server computer, and includes a storage section DB1, a calculation section CPU, an input section 30, and an output section .

The storage unit DB1 stores incoming/outgoing/inventory information D1, predicted parts information D2, past item information D3, similar specification information D4, unified specification information D5, new/past item information D6, and specification group information D7. , correlation degree information D8, score information D9, specification selection result information D10, prediction target case information D11, and model prediction information D12.

These pieces of information can be grasped by dividing them into pre-stored initial information, final product information which is the calculation result, and intermediate product information obtained during the calculation. The initial information includes incoming/outgoing/inventory information D1, past item information D3, and prediction target item information D11, which are indicated by dotted lines in FIG. The final product information is the predicted parts information D2 and the type prediction information D12, which are indicated by thick solid lines in FIG. Information other than this is intermediate product information, and is indicated by a thin solid line in FIG. The details of these pieces of information will be explained in subsequent descriptions.

The processing unit CPU functionally expresses the processing contents executed here by an information acquisition unit 11, a target part selection unit 12, a specification grouping unit 13, a correlation measurement unit 14, and a specification selection unit 15. , and the model prediction unit 16. FIG. Only an outline of each function is described here.

First, the information acquisition unit 11 obtains from the storage unit DB1 information necessary for processing in the target part selection unit 12, the specification grouping unit 13, the correlation measurement unit 14, the specification selection unit 15, and the model prediction unit 16. is obtained, and the processing result is stored in the storage unit DB1.

At the start of the process, the storage unit DB1 stores the incoming/outgoing/inventory information D1, the past item information D3, and the prediction target item information D11 as the initial information indicated by thin lines in FIG.

The target parts selection unit 12 extracts parts that may be out of stock or overstock from all parts of the elevator from the warehousing/shipping/inventory information D1, and stores the processing result in the storage unit DB1 as predicted parts information D2. Store.

The specification grouping unit 13 analyzes the similarity of each specification item extracted from the past project information D3, which is the initial information, based on the predicted parts information D2, and then stores the analyzed data as similar specification information D4 in the storage unit DB1. Furthermore, by using the similar specification information D4, the item names of the specifications having the same content but different notation are unified, and stored in the storage unit DB1 as the unified specification information D5. Thereafter, based on the unified specification information D5, the specification items of the past case information D3 are unified and stored in the storage unit DB1 as new past case information D6. Next, specifications are classified into a plurality of groups from the new/past case information D6 based on the input tendency of specifications, and stored in the storage unit DB1 as specification group information D7.

The correlation degree measuring unit 14 sorts the specifications in descending order of the degree of correlation between the specifications and the model type for each group of the specification group information D7, and stores them in the storage unit DB1 as the correlation degree information D8. Also, in order to delete specifications with a low frequency of use, the number of past cases corresponding to each specification is totaled using the new past case information D6, and stored as the number of past cases in the correlation degree information D8 in the storage unit DB1.

The specification selection unit 15 extracts the combination of specifications for each group in descending order of correlation from the correlation degree information D8, excluding specification information with a small number of past cases. A prediction model is learned from the new past case information D6, and the prediction accuracy of the learned model is stored in the storage unit DB1 as score information D9. Also, from the score information D9, a combination of specifications with the smallest score is extracted for each group and stored in the storage unit DB1 as specification selection result information D10.

The model prediction unit 16 learns a prediction model for each specification group from the specification selection result information D10 and the new/past item information D6, and predicts the part type to be predicted for each item in the prediction target item information D11. The result is stored in the storage section DB1 as the model prediction information D12.

The input unit 30 is connected to the user terminal 50 and the database DB2 via the network Nw, receives the incoming/outgoing/inventory information D1, the past item information D3, and the prediction target item information D11 from the database DB2, and stores the storage unit DB1. store to As a result, the initial information necessary for executing subsequent processing is secured in the storage unit DB1.

The output unit 40 outputs the predicted parts information D2 stored in the storage unit DB1, the specification group information D7, the correlation degree information D8, the score information D9, the specification selection result information D12, and the model prediction information D12 to the user terminal. 50 and display the results to the user.

Next, the flow of processing executed by the component type prediction device 1 in this embodiment will be described with reference to the flowchart of FIG. Note that the following series of processes are based on the premise that the incoming/outgoing/inventory information D1, the past case information D3, and the prediction target case information D11 are recorded in the external database DB2. is started in response to a start command from the user.

In response to a start command from the user, the information acquisition unit 11 acquires the incoming/outgoing/inventory information D1 and the past item information D3 from the database DB2 via the network Nw, and stores them in the storage unit DB1. As a result, the initial information necessary for executing subsequent processing is secured in the storage unit DB1.

In such a state, in processing step S1, the target part selection unit 12 selects parts that may be out of stock or in excess inventory as target parts from the incoming/outgoing/inventory information D1. The warehousing/shipping/inventory information D1, which is the initial information used at this time, includes the data illustrated in FIG.

FIG. 3 shows an example of the data structure of the incoming/outgoing/inventory information D1. In the warehousing/delivery/inventory information D1, the information of the warehousing record D1d, the warehousing record D1e, the inventory record D1f, and the safety stock D1g of all parts of the elevator is stored together with the part D1a, the model D1b, and the date D1c.

Here, the part D1a indicates number information for identifying the part. The model D1b indicates the part model number that constitutes the part. The date D1c indicates the date on which the warehousing/delivery/inventory results are aggregated. The warehousing amount D1d indicates the actual number of warehousing of parts. For example, 100 units of “type 1” in FIG. The shipping amount D1e indicates the actual number of parts shipped, and for example, "type 1" in FIG. 2 indicates that 300 units were shipped in "21/1". The inventory amount D1f indicates the actual number of parts in inventory. For example, the "model 1" in FIG. The safety stock quantity D1g indicates the safety stock quantity of parts, and for example, the safety stock quantity of “model 1” in FIG. 3 is 300 units.

In the determination of the shortage or excess inventory in the processing step S2, for example, if the model D1b in FIG. 40 units, which is less than "300 units" of the safety stock D1g. In this way, it is determined that there is a possibility that the part whose stock quantity D1f is less than the safety stock quantity D1g is out of stock. Similarly, when the model D1b in FIG. It can be seen that there are more than 100 units. In this way, it is determined that there is a possibility of excess inventory for parts whose inventory quantity D1f exceeds the safety inventory quantity D1f. Therefore, model 1 and model 2 are treated as target parts, and are stored in storage unit DB1 as predicted parts information D2 in FIG. In the following, processing steps S2 to S9 are repeated for each component to be handled selected in processing step S1.

FIG. 4 is an example of the data structure of the predicted parts information D2 created in processing step S1. The predicted parts information D2 stores the results of extracting parts that are likely to be out of stock or overstocked from the parts information registered in the warehousing/shipping/inventory information D1 of FIG. and part type D2b. The prediction target part D2a indicates the name of the prediction target part. The part type D2b indicates the type of the part. For example, when the prediction target part D2a in FIG. .

Next, in processing step S2 of the flow of FIG. 2, past case information D3, which is initial information, is read. FIG. 5 is an example of the data structure of the past case information D3.

In the past case information of FIG. 5, information related to past cases of parts registered in the predicted parts information D2 is stored. Elevator specification information such as D3f, weight D3g, speed D3h, speed D3i, door width D3j, opening direction D3k, door type D3l, part name D3m, and model name D3n.

Of these, the case D3a indicates number information for identifying the case. The customer name D3b indicates the customer name of each item. The delivery date D3c indicates the deadline for delivering the product to the customer.

Loaded weight D3d, weight D3e, loaded D3f, and weight D3g indicate the loaded weight of the product desired by the customer. For example, if case D3a in FIG. 5 is "1", the customer wants a product that can be loaded up to 500 kg. Regarding the loading weight, the loading weight D3d, weight D3e, loading D3f, and weight D3g are notated because there are many cases where the specifications are the same but the notation is inconsistent. Even if there is shaking, it is possible to uniquely grasp the specifications.

Speed D3h and speed D3i indicate the speed of the product desired by the customer. For example, if the case D3a in FIG. 5 is "1", the customer wants a product with a speed of 120 m/min. The door width D3j indicates the door width of the product desired by the customer. For example, if case D3a in FIG. 5 is "1", the customer wants a product with a door width of 1000 mm. The opening direction D3k indicates the opening direction of the product desired by the customer. For example, if the case D3a in FIG. 5 is a case of "2", the customer desires a product with an L opening direction. The door type D3l indicates the product type of the constituting door. For example, if the case D3a in FIG. 5 is "1", the customer wants a type 1 door. The part name D3m indicates the target part name. For example, if case D3a in FIG. 5 is "1", the part name constitutes part 1. The model name D3n indicates the target part model. For example, if case D3a in FIG. In addition, in each data item, when data is not entered, it is indicated as "-".

Next, in processing step S3 of the flow of FIG. 2, grouping is performed for each specification. This processing is because the specification information used for each item differs depending on the model ordered, etc., and therefore, it is necessary to divide the specifications into a plurality of groups.

Therefore, in processing step S3, after the specification grouping unit 13 extracts the specification items of the past case information D3, the similarity of each extracted specification item is analyzed using natural language processing or the like as the first process, Similar specification items and similarity scores are stored in the storage unit DB1 as similar specification information D4. Further, as a second process, in the similar specification information D4, for example, similar specifications with a degree of similarity of 0.5 or more are grouped as specifications with different notations but with the same content, and stored in the storage unit DB1 as unified specification information D5. Further, as the third process, the standardized specification information D5 is used to unify specification items with the same contents but different notations in the past case information D3, and the processing result is stored in the storage unit DB1 as the new past case information D6. As the fourth process, in the post-new/past case information D6, the specification information is grouped from the tendency of specifications to be used, and the storage part DB1 is stored as the processing result in the specification group information D7.

Regarding the first process, specifically, first, the specification grouping unit 13 performs specification items (loaded weight D3d, weight D3e, loaded D3f, weight D3g, speed D3h, speed D3i, door width D3j, opening direction D3k, door type D3l), similar specification items are analyzed, and the similar specifications and the degree of similarity are stored in the storage unit DB1 as similar specification information D4 in FIG.

FIG. 6 is an example of the data structure of the similar specification information D4. The similar specification information D4 stores specifications similar to each specification item registered in the past case information D3, and is composed of a specification item D4a, a similar specification D4b, and a score D4c. The specification item D4a indicates the specification item name registered in the past case information D3. For example, the specification item D4a of the similar specification information D4 in FIG. 6 indicates data when the specification item of the past case information D3 in FIG. 5 is "load weight D3d". The similar specification D4b indicates specification items other than the specification items of the similar specification information D4 among the specification items registered in the past item information D3. For example, when the specification item D4a of the similar specification information D4 of FIG. 6 is "load weight", the similar specification D4b is a specification item other than the load weight among the specification items of the past case information D3 of FIG. Load D3f, weight D3g, speed D3h, and the like. The score D4c numerically indicates the degree of similarity between each specification item D4a and the similar specification D4b. For example, when the specification item of the similar specification information D4 in FIG. 6 is "load weight" and the similar specification is "weight", the score indicates that the similarity between these two specifications is 0.98.

For example, when the specification item D4a is "load weight", similarity with other specification items D4a such as "weight", "load", "weight", and "door width" is calculated using cosine similarity. It analyzes and stores the processing result in the storage unit DB1 as the similar specification information D4. Here, when the specification item of the similar specification information D4 in FIG. is "0.83".

Next, regarding the second process, the similar specification information D4 in FIG. 6 is used to unify the specification items with the same content but with different specification descriptions, and the processing result is stored in the storage unit DB1 as the unified specification information D5 in FIG. do.

FIG. 7 is an example of the data structure of the unified specification information D5. The unified specification information D5 consists of pre-unification D5a and post-unification D5b. For example, when a similar specification D4b having a similarity (score) D4c of 0.5 or more in the similar specification information D4 in FIG. The specifications whose item D4a has the same content as "loaded weight" are "weight", "load", and "weight".

After that, in the unified specification information D5 of FIG. 7, the grouped similar specifications "loading weight", "loading", "weight", and "weight" are transferred to the pre-unification D5a, which is the first analyzed specification item "loading Weight” is stored in D5b after unification. Here, the analysis of similar specifications proceeds in the order of the specification items of the past case information D3 in FIG. and

For example, if the specification item of the past case information D3 in FIG. 4 is "weight", the specification item is grouped as similar specifications when analyzing "loading weight", and the "before unification" column of the unified specification information D5 in FIG. Since it is already stored in D5a, analysis is not performed, and analysis of the next specification item "speed" is started.

Next, regarding the third process, after that, using the unified specification information D5 of FIG. 7, the notation of the specification items of the past case information D3 of FIG. to store

FIG. 8 shows an example of the data structure of the new past case information D6. The new past case information D6 stores the result of correcting the specification information of the past case information D3 based on the standardization of the specification notation of the unified specification information D5. , a load weight D6d, a speed D6e, a door width D6f, an opening direction D6g, a door type D6h, a part name D6i, and a model name D6j.

The contents of each item are the same as the items of the past case information D3. For example, in the case of the new past case information D6 in FIG. 8, the specifications notation of "loading weight", "weight", "loading", and "weight" in the past case information D3 in FIG. ”, and “speed” are unified into “speed”, and all the columns before unification are aggregated into the column after unification.

For example, in the past case information D3 in FIG. 5, if the case is "1", the specification items "500 kg" in weight, "500 kg" in loading, and "500 kg" in weight are specified in the unified specification information D5 in FIG. Since it is "loading weight" after unification, the notation is unified to "loading weight", and the processing result is stored in the storage unit DB1 as the new past case information D6.

Furthermore, regarding the fourth process, using the new past case information D6, specification information with similar tendencies of used specifications is classified by a solution method that can classify target characteristics from past tendencies such as other class classification. , the classification result is stored in the storage unit DB1 as the specification group information D7.

FIG. 9 is an example of the data structure of the specification group information D7. The specification group information D7 stores the result of classifying the specifications into a plurality of groups in the new past case information D6 of FIG. 8, and is composed of a specification group D7a and specification information D7b. The specification group D7a indicates a specification group name that identifies a group of specification information. The specification information D7b indicates specification information belonging to a specification group. For example, when the specification group in FIG. 9 is "group 1", the belonging specification information is "load weight", "speed", "door width", and "door type".

For example, if the item is "1" in the new/past item information D6 in FIG. 8, there is no description of the opening direction, so the used specifications are "load weight/speed/door width/door type". In addition, when the case is "2" or "3", since there is no description of the door width, the specifications to be used are "load weight, speed, opening direction, and door type." Based on the tendency of the specifications to be used in this way, we divided them into specification group 1 for "load weight, speed, door width, door type" and specification group 2 for "load weight, speed, opening direction, door type". It is stored in the storage unit DB1 as the specification group information D7.

Returning to FIG. 2, in processing step S4, the correlation measuring unit 14 measures the degree of correlation between the specification and model for each specification group from the new/past item information D6 and the specification group information D7, and stores it as correlation degree information D8. Store in part DB1.

FIG. 10 is an example of the data structure of the correlation information D8. Correlation degree information D8 stores the result of analyzing the degree of correlation between the specification information of new/past item information D6 and the part type for each specification group of specification group information D7. , the degree of correlation D8c, and the number of past cases D8d. The specification group D8a indicates a specification group name that identifies a group of specification information D8a. The specification information D8a indicates specification information belonging to the specification group. The degree of correlation D8c indicates the degree of correlation between the specification information in the new past case information D6 and the model to be predicted.

For example, when the specification group D8a in FIG. 10 is "group 1" and the specification information D8b is "loaded weight", the degree of correlation D8c is 0.97. The number of past cases D8d indicates the number of cases of the corresponding specifications in the new past case information D6. For example, if the specification group in FIG.

Specifically, for example, when the specification group information D7 in FIG. 9 is "group 1", the specification information is "load weight", "speed", "door width", and "door type". From the new past case information D6 in FIG. 8, the degree of correlation between the specification information and the model to be predicted is calculated using Cramer's correlation coefficient or the like, and is stored in the storage unit DB1 as correlation degree information D8. Also, using the new past case information D6, the number of cases in which each specification information is used is totaled and stored in the storage unit DB1 as the past case number D8d of the correlation degree information D8.

Next, in processing step S5 in FIG. 2, the information acquisition unit 11 acquires the latest prediction target case information D11 from the database DB2 via the network Nw, and stores it in the storage unit DB1. Next, the specification selection unit 13 repeats the processing from processing steps S6 to S7 until the score generated in processing step S7, which will be described later, is no longer improved.

In the repeated processing of processing steps S6 to S7, first, in processing step S6, the specification information selection unit 13 extracts specification combination candidates for each specification group from the correlation degree information D8, and processes them into the storage unit DB1 as score information D9. Store the result.

Specifically, first, specifications with a low frequency of use are excluded from the correlation degree information D8 using a non-correlation test or the like. For example, when the specification group of the correlation information D8 in FIG. 9 is "group 1", the specification information with the highest correlation is "load weight", "speed", "door width", and "door type" in order from the top. becomes. As a result of using the non-correlation test, if the number of projects required in the prediction is 10, the number of past projects with the specification information "speed" is 8, so this specification is excluded. After excluding specifications with a low frequency of use in this way, specification information with a high degree of correlation is combined in order from the top, and the result is stored in the storage unit DB1 as score information D9. For example, in the case of the above example, if the combination of specifications is set in descending order of the degree of correlation, then "load weight", "load weight/door width", and "load weight/door width/door type" are set. Combination candidates of these specifications are stored in the storage unit DB1 as score information D9.

In the processing step S7 of the repetitive process, the specification selection unit 13 learns the model prediction model from the new past case information D6 and the score information D9 in order from the top of the specification combination candidates, calculates the prediction error of the prediction model, The result is stored in the storage section DB1 as the score information D9.

FIG. 11 is an example of the data structure of the score information D9. The score information D9 stores the measurement result of the prediction accuracy when a prediction model is learned by combining specifications with a high degree of correlation for each specification group from the correlation information D8 and the new/past item information D6. It consists of a group D9a, a specification combination D9b, and a score D9c.

The specification group D9a indicates a specification group name that identifies a group of specification information. The specification combination D9a indicates the result of combining specification information with a high degree of correlation for each specification group from the correlation degree information D8. The score D9c indicates the prediction error when the prediction model is learned using the new past case information D6 based on the specification information of the specification combination, and the smaller the error, the higher the accuracy of the prediction model. For example, when the specification group in FIG. 11 is "group 1" and the specification combination is "load weight", the score is 0.431. If the specification group in FIG. 11 is "group 1" and the specification combination is "load weight/door width", the score is 0.282. If the specification group in FIG. 10 is "group 1" and the specification combination is "load weight/door width/door type", the score is 0.644.

Specifically, for each of the specification combination candidates calculated in the processing step S6, the specification information is used as an explanatory variable and the type as an objective variable, and a prediction model of the new past case information D6 is learned using multiple regression analysis or the like. After the prediction model is learned, the loss function of the prediction model such as the square sum error between the predicted value and the actual value is stored in the storage unit DB1 as the score information D9. For example, when the specification combination of the score information D9 in FIG. 11 is "loaded weight", the score of the loss function is "0.431", so "0.431" is stored in the score.

Finally, in FIG. 2, processing steps S6 and S7 are repeated for each specification group until the score of the combination of specifications does not improve. For example, if the specification group of the score information D9 in FIG. Minimal pattern. Therefore, the iteration ends when the score for the next combination of specifications, ie, "load weight/door width/door type" is calculated.

In FIG. 2, in processing step S8, the specification selection unit 13 extracts the combination of specifications with the smallest score for each specification group from the score information D9, and stores the combination in the storage unit DB1 as specification selection result information D10. For example, if the specification group of the score information D9 in FIG. 11 is "group 1", the combination of specifications with the lowest score is "load weight/door width". Store.

FIG. 12 is an example of the data structure of the specification selection result D10. The specification selection result D10 stores the combination of specifications with the lowest score among the specification combinations obtained from the score information D9 for each specification group. D10b. The specification group D10a indicates a specification group name that identifies a group of specification information. The specification D10b used for prediction indicates the combination of specifications with the lowest score among the specification combinations obtained from the score information D9. For example, when the specification group in FIG. 11 is "Group 1", the specifications used in the prediction are load weight and door width.

In FIG. 2, in processing step S9, the information acquisition unit 11 acquires prediction target case information D10 from the database DB2 via the network Nw, and stores it in the storage unit DB1.

FIG. 13 shows an example of the data structure of the prediction target case information D11. Prediction object item information D11 stores registration information of current order items, including item D11a, customer name D11b, delivery date D11c, loading weight D11d, weight D11e, loading D11f, weight D11g, speed D11h. , speed D11i, door width D11j, opening direction D11k, and door type D11l.

In subsequent processing step S9, the model prediction unit 16 learns a model prediction model for each specification group from the prediction target project information D11, the new/past project information D6, and the specification selection result information D10. Predict the type used. Then, the processing result is stored in the storage unit DB1 as the model prediction information D12.

FIG. 14 is an example of the data structure of the model prediction information D12. The model prediction information D12 stores the result of predicting the model to be predicted for each project in the prediction target project information D11 after building a prediction model for each group from the specification selection result D10 and the new/past project information D6. It consists of a case D12a, a specification group D12b, a specification D12c, a part D12d, and a model prediction D12e. The case D12a indicates number information for identifying a prediction target case. The specification group D12b indicates the specification group of the project to be predicted. For example, if case D12a in FIG. 14 is "1", specification group D12b belongs to group 1. The specification D12c indicates a combination of specifications used when constructing a prediction model. For example, when the case D12a in FIG. 13 is "1" and the specification group D12b is "group 1", the specification D12c is load weight and door width. The part D12d indicates a prediction target part for each case. The predicted type D12e indicates the result of predicting the type of parts used in each case. For example, when case D12a in FIG. 13 is "1", part D12e is part 1 and the predicted model is model 1.

For example, when the specification group of the specification selection result information D12 of FIG. 12 is "group 1", the specification used for prediction is "load weight/door width". Next, from the new past project information D6 in Fig. 8, the past projects whose specification group belongs to "group 1" are extracted, and the relationship between the specification information "load weight/door width" and the model is determined using multiple regression analysis. learn. The result of later prediction is stored in the storage unit DB1 as model prediction information D12.

The information as the final product obtained as described above is externally output via the output unit 40 of FIG. constructed and presented to the user through a monitor.

FIG. 15 shows a configuration example of a monitor output screen 400 in this case. The illustrated output screen 400 displays a target part selection result 401, a specification selection result 402, and a model prediction result 403, which are information as final products.

Here, the target part selection result 401 is displayed in list form, and the list is formed of a part column 151 , a model column 152 , a risk column 153 and a selection result column 154 . The specification selection result 402 is composed of an item button 155 for selecting an item and a list of the specification selection results of the selected item. It consists of a number column 159 and a specification selection result column 160 . The model prediction result 403 list is composed of a parts column 161 and a predicted model column 162 .

In this display screen, the names of all the parts of the elevator are displayed in the part column 151 . The model column 152 displays the model name of each component. A risk column 153 displays whether or not there is a possibility that the model belonging to the model column 152 is out of stock or in excess stock. The selection result column 154 displays whether or not the item is likely to be out of stock or overstocked in the risk column 153 described above. The matter number column 155 displays information about the matter number specified by the user. A specification group is displayed in the specification group column 156 . The specification information column 157 displays the specification information belonging to the specification group column 156 described above. The degree of correlation field 158 displays the degree of correlation between the specification information field 157 and the model. The number of past cases column 159 displays the number of past cases using the specification information column 157 described above. In the specification selection result column 160, the determination result of whether or not to use the specification information column 157 for model prediction is displayed. The part column 161 displays the name of the part when the model is predicted based on the specification selection result column 160 described above. The predicted model column 162 displays the model name when the model is predicted based on the specification selection result column 160 described above.

The present invention has been described in detail based on the embodiments above. It predicts the type of project. In addition, when making predictions, in more detail, past projects are read, specifications that are candidates for model prediction are selected from a huge number of specifications, and specifications to be used for prediction are selected from the narrowed down candidates. , builds a prediction model, and uses the specification information to predict the type of parts that have not yet been installed in the latest order.

In order to realize this, it is necessary to select specifications that are candidates for model prediction from a huge number of specifications, select specifications to be used for prediction from the narrowed down candidates, and build a prediction model. The important thing is how to realize it.

For this reason, the selection of specification candidates is divided into pre-processing and main processing. elements are standardized. For example, even with the same specifications, there are variations in notation depending on the designer. Therefore, natural language processing is used to unify overlapping specification information. Because it is human and not resilience, by analyzing the combination of specifications used, it is classified into multiple groups that do not depend on the designer. In this process of selecting specification candidates, after pre-processing, specifications with low frequency of use, such as specifications unique to the project, are excluded for each group, and specification candidates with a high degree of correlation with the model to be predicted are selected.

As for constructing the prediction model, the combination of specifications to be used for prediction is extracted in order from the specification with the highest degree of correlation for each group, and the prediction model is learned repeatedly until the accuracy of the model is minimized.

As a result, when compared with the conventional method, the present invention learns the relationship between the specification information and the model using the past data, and builds a model prediction model to identify the model that has not been completed. By predictive model learning, even if specifications are added or deleted, it can be handled. Moreover, since the specification information necessary for prediction is extracted in consideration of correlation, the prediction accuracy can be increased.

In addition, in the present invention, since the model of parts is predicted within the range of specifications for which installation has been completed without using information on specifications for which installation has not been completed, the model is directly predicted from the specifications for which installation has been completed, resulting in high accuracy. In addition, even if the specifications are enormous, it can be handled.

As described above, according to the present invention, after selecting long-delivery parts that have the risk of missing parts or excess inventory, specifications that are highly correlated with the model are selected from a huge amount of specification information, and high-precision Models can be predicted.

1: Part model prediction device 11: Target part selection unit 12: Information acquisition unit 13: Specification grouping unit 14: Correlation degree measurement unit 15: Specification selection unit 16: Model prediction unit 30: Input unit 40: Output unit 50: User terminal 400: Output screen CPU: Calculation unit DB2: Database DB1: Storage unit Nw: Network 30: Input unit 40: Output unit D1: Warehousing/shipping/inventory information D2: Predicted parts information D3: Past item information D4: Similar specification information D5 : Unified specification information D6: New and past item information D7: Specification group proposal information D8: Correlation degree information D9: Score information D10: Specification selection result information D11: Prediction target item information D12: Model prediction information

Claims (10)

a storage unit that stores warehousing/shipping/inventory information that records past warehousing/shipping/inventory numbers of parts that make up a product and past item information that records specifications/models of past items;
a target parts selection unit that selects parts to be predicted from the incoming/outgoing/inventory information;
a specification grouping unit that classifies specifications to be used into a plurality of groups from the past case information;
a correlation degree measuring unit that extracts specifications with a high frequency of use from the specification group information and calculates the degree of correlation between the specifications and the model;
a specification selection unit that selects specifications required for prediction by combining specifications with a high degree of correlation for each group from the correlation information and the prediction target project information;
A parts model prediction device comprising a model prediction unit for predicting a target model from specification selection result information. The part type prediction device according to claim 1,
The storage unit stores incoming/outgoing/inventory information, predicted parts information, past item information, similar specification information, unified specification information, new/past item information, specification group information, correlation information, and scores. A part model prediction device, characterized by storing information, specification selection result information, and model prediction information. The part type prediction device according to claim 1,
The part model prediction device, wherein the target part selection unit selects a part that is highly likely to be out of stock or in excess inventory based on past receipts, deliveries, and inventory transitions of elevator parts. The part type prediction device according to claim 1,
The part type prediction device, wherein the specification grouping unit unifies specification information having the same content but different notations from past project information, and classifies the specifications into a plurality of groups based on the used specification information. The part type prediction device according to claim 1,
The correlation degree measuring unit calculates the candidate specifications by rearranging the specifications in descending order of the degree of correlation between the specifications and the model in the specifications with high frequency of use for each group of the specification group information. prediction device. The part type prediction device according to claim 1,
The specification selection unit sets a combination of specifications in descending order of correlation for each group from the correlation information, repeatedly learns the prediction model until the score does not improve, and selects the specifications necessary for prediction. Characterized part type prediction device. The part type prediction device according to claim 1,
wherein the model prediction unit learns a prediction model from specification selection result information, specification group information, past project information, and target component information, and predicts the model of the project to be predicted. . When reading order information and predicting the model of an unfinished project based on specification information and past shipping information, read the past project and specify the specifications that are candidates for predicting the model from among multiple specifications. is selected, the specifications to be used for prediction are selected from among the narrowed down candidates, a prediction model is constructed, and the specification information is used to predict the uncompleted part type in the latest order. Part type prediction method. The part type prediction method according to claim 8,
When selecting specifications that are candidates for model prediction from among multiple specifications, we standardize individual factors and exclude specifications that are used infrequently for each group, and have a high degree of correlation with the model to be predicted. A parts type prediction method, characterized by selecting candidates for specifications. The part type prediction method according to claim 8,
Select the specifications to be used for prediction from the narrowed down candidates, and when constructing the prediction model, extract the combination of specifications to be used for prediction in order from the specifications with the highest degree of correlation for each group, and minimize the accuracy of the model. A part type prediction method characterized by learning a prediction model iteratively up to

Images