TWI877796B - Driver inclination prediction device, driver inclination prediction method, learning device, learning method, and information processing program - Google Patents

Abstract

[Topic] Provide a mechanism for predicting a driver with a predetermined inclination based on characteristics of the driver himself. [Solution] An information processing device obtains user characteristics representing characteristics of a target user from one or more first services; obtains vehicle characteristics representing characteristics of a vehicle owned by the target user from a second service different from the one or more first services; and predicts whether the target user is a driver with the predetermined inclination by using a learning model that takes the user characteristics and the vehicle characteristics as inputs and outputs the probability that the target user is a driver with the predetermined inclination after learning.

Description

Driver inclination prediction device, driver inclination prediction method, learning device, learning method, and information processing program

The present invention relates to a driver inclination prediction device, a driver inclination prediction method, a learning device, a learning method, and an information processing program, and more particularly to a technology required by a driver for predicting a predetermined inclination.

Drivers of vehicles (including cars and bicycles) are required to drive safely. Drivers who are judged to be able to drive safely are considered to be drivers with good tendencies. Once a driver is considered to be a driver with good tendencies, he or she may receive various preferential treatments. For example, if a driver with good tendencies purchases automobile insurance for his or her car, the insurance premium may be reduced to a lower amount than that of a driver without good tendencies. Previously, whether a driver was a good driver was determined based on physical data obtained from the vehicle used or whether the driver had caused any accidents.

For example, Patent Document 1 discloses that the driver's driving proficiency is determined by evaluating the engine speed change rate, the rate of sudden braking, and the rate of sharp turns in the car used by the driver. According to this document, a driver with high driving proficiency can be determined as a driver with excellent tendencies. [Prior Technical Document] [Patent Document]

[Patent Document 1] Japanese Patent Application Publication No. 2014-65362

[The problem that the invention wants to solve]

As mentioned in the above document, whether a driver has a predetermined tendency such as good tendency has been judged based on physical data obtained through driving or objective data such as the presence or absence of accidents. However, a mechanism for predicting whether a driver has a predetermined tendency based on the characteristics of the driver himself such as the driver's life background rather than such objective data has not been proposed so far. If a driver with a predetermined tendency can be predicted based on the characteristics of the driver himself, the prediction can be made even if there is no objective data about the driver.

The present invention was developed in view of the above-mentioned problem, and its purpose is to provide a mechanism required by the driver to predict a certain tendency based on the characteristics of the driver himself. [Means for solving the problem]

In order to solve the above-mentioned problem, according to one aspect of the driver tendency prediction device of the present invention, it has: user characteristic acquisition means for acquiring user characteristics representing characteristics of the target user from one or more first services; and vehicle characteristic acquisition means for acquiring vehicle characteristics representing characteristics of the vehicle owned by the previous target user from a second service different from the one or more first services; and prediction means for predicting whether the previous target user is a driver with the previously specified tendency by using a learning model that takes the previous user characteristics and the previous vehicle characteristics as inputs and outputs the probability that the previous target user is a driver with the specified tendency after learning. In order to solve the above-mentioned problem, according to one aspect of the learning device of the present invention, it has: user feature acquisition means, which is used to acquire user features representing features of a plurality of users from one or more first services; and vehicle feature acquisition means, which is used to acquire vehicle features representing features of vehicles owned by the aforementioned plurality of users from a second service different from the aforementioned one or more first services. service, and a level acquisition means for acquiring the levels of vehicle insurance of vehicles owned by a plurality of users; and a generation means for generating learning data based on the aforementioned user characteristics, the aforementioned vehicle characteristics, and the aforementioned levels; and a learning means for using the aforementioned learning data to enable a learning model required for predicting the probability of a driver with a given tendency to learn.

In order to solve the above-mentioned problem, one aspect of the driver tendency prediction method according to the present invention comprises: a user characteristic acquisition step of acquiring user characteristics representing characteristics of a target user from one or more first services; and a vehicle characteristic acquisition step of acquiring vehicle characteristics representing characteristics of a vehicle owned by the target user from a second service different from the one or more first services; and a prediction step of predicting whether the target user is a driver with the predetermined tendency by using a learning model that takes the target user characteristics and the vehicle characteristics as inputs and outputs the probability that the target user is a driver with the predetermined tendency after learning. In order to solve the above-mentioned problem, according to one aspect of the learning method of the present invention, there are: a user characteristic acquisition step, which is to acquire user characteristics representing characteristics of a plurality of users from one or more first services; and a vehicle characteristic acquisition step, which is to acquire vehicle characteristics representing characteristics of vehicles owned by the plurality of users from a first service different from the one or more first services. 2 service, and obtain it; and the level acquisition step is to obtain the vehicle insurance levels of the vehicles owned by the plurality of users; and the generation step is to generate learning data based on the user characteristics, the vehicle characteristics, and the levels; and the learning step is to use the learning data to learn the learning model required for predicting the probability of a driver with a given tendency.

In order to solve the above-mentioned problem, according to one aspect of the information processing program of the present invention, the information processing program is used to make the aforementioned computer execute the following processing: user characteristic acquisition processing, which is to acquire user characteristics representing characteristics of the target user from one or more first services; and vehicle characteristic acquisition processing, which is to acquire information representing characteristics of the aforementioned target user from one or more first services; Vehicle characteristics such as characteristics of a vehicle owned by a user are obtained from a second service different from one or more first services mentioned previously; and prediction processing is to predict whether the previous target user is a driver with the previously specified tendency by using a learning model that takes the previous user characteristics and the previous vehicle characteristics as inputs and outputs the probability that the previous target user is a driver with the specified tendency after learning. Another aspect of the information processing program according to the present invention is an information processing program for causing a computer to execute information processing, and the program is used to cause the aforementioned computer to execute the following processing: user characteristic acquisition processing, which is to acquire user characteristics representing characteristics of a plurality of users from one or more first services; and vehicle characteristic acquisition processing, which is to acquire vehicle characteristics representing characteristics of vehicles owned by the aforementioned plurality of users from one or more first services. A second service different from the first service mentioned above, which is one or more, is obtained; and a level acquisition process is to obtain the levels of vehicle insurance of vehicles owned by the plurality of users mentioned above; and a generation process is to generate learning data based on the user characteristics mentioned above, the vehicle characteristics mentioned above, and the levels mentioned above; and a learning process is to use the learning data mentioned above to make the learning model required for predicting the probability of a driver with a given tendency to learn. [Effect of the invention]

According to the present invention, a mechanism for predicting a driver's predetermined inclination based on the characteristics of the driver himself can be provided. The above-mentioned purpose, aspect and effect of the present invention and the purpose, aspect and effect of the present invention not mentioned above can be understood by the person skilled in the art by referring to the attached drawings and the description of the patent application scope, and the following forms required for implementing the invention can be understood.

The following detailed description of the implementation forms required for implementing the present invention is made with reference to the attached drawings. Among the components disclosed below, those having the same function are marked with the same symbols, and their descriptions are omitted. In addition, the implementation forms disclosed below are used as an example of the means for implementing the present invention, and must be appropriately modified or changed according to the structure or various conditions of the device to which the present invention is applied. The present invention is not limited to the following implementation forms. In addition, all combinations of the features described in this implementation form are not necessarily necessary for the solution of the present invention.

[Configuration of information processing system] FIG. 1 illustrates an example configuration of an information processing system according to the present embodiment. As an example, the present information processing system, as shown in FIG. 1, includes: an information processing device 10, a plurality of user devices 11-1 to 11-N (N>1) used by any plurality of users 1 to N, and a user device 11-T1 used by a target user T1. In addition, in the following description, unless otherwise specified, user devices 11-1 to 11-N and 11-T1 are collectively referred to as user devices 11. In addition, in the following description, the terms user device and user can be used synonymously.

The user device 11 is a device such as a smart phone or a tablet, and is configured to communicate with the information processing device 10 through a public network such as LTE (Long Term Evolution) or a wireless communication network such as a wireless LAN (Local Area Network). The user device 11 has a display unit (display surface) such as a liquid crystal display, and each user can perform various operations through the GUI (Graphic User Interface) equipped in the liquid crystal display. The operation includes: various operations on the content of the image displayed on the screen, such as touch operations, sliding operations, and scrolling operations performed with fingers or styluses. In addition, the user device 11 is not limited to the device of the form shown in FIG. 1, and can also be a desktop PC (Personal Computer) or a notebook PC. In this case, the operations performed by each user can be performed using an input device such as a mouse or a keyboard. In addition, the user device 11 may also have a display surface.

The user of the user device 11 can use a plurality of Web services (Internet-related services) provided by the information processing device 10 or other devices (not shown) through the information processing device 10. These Web services can be used by the user by, for example, registering to use the API (Application Programming Interface) provided by the information processing device 10 or other devices.

In this embodiment, the Web service includes a vehicle service. In this embodiment, the vehicle service is a service that provides services related to cars. For example, the vehicle service provides users with information about gas stations or preferential information about car supplies. In addition, the vehicle service may include a service that gives points to users or issues coupons to users by allowing them to register information about their own cars. The information about the car may include information such as the manufacturer name, car name, model, and replacement frequency of the car.

Furthermore, in this embodiment, the Web service includes a car insurance service (car insurance service). The car insurance service is a service that allows users who have registered for the service to receive compensation when an accident occurs due to a car owned by the user by paying an amount (insurance premium) corresponding to the assigned level. The level is assigned by the insurance company that operates the car insurance service, and a high level is assigned to a user who is judged to be a safe driver. Therefore, the higher the level, the higher the discount rate (or the higher the reward rate), and the lower the amount paid by the user. This level is assigned to you when you start using the service, and you will be assigned a higher level after a certain period of time by meeting certain conditions such as no incidents.

Furthermore, the web service may include an online shopping mall or online supermarket, or services related to communications, finance, real estate, sports, and travel. The user device 11 can transmit information about the user of the user device 11 to the information processing device 10 by utilizing such a web service. For example, the user device 11 can transmit information about the characteristics of the user device or the user, such as the IP (Internet Protocol) address of the user device 11, the user's address, or the user's name (first and last name), to the information processing device 10. Furthermore, the user device 11 can perform positioning calculations based on signals received from a GPS (Global Positioning System) satellite (not shown), etc., and the information obtained by the calculation is generated as the location information of the user device 11 and transmitted to the information processing device 10.

In addition, in this embodiment, the plurality of Web services that can be used by the user of the user device 11 can be directly or indirectly managed and operated by a specific management company and its group company (affiliated companies of the management company). Furthermore, the plurality of Web services are configured so that information (such as information about the user such as user characteristics) can be managed using a user ID that is common among the plurality of Web services. For example, electronic credits such as points that are common among the plurality of Web services can be given to the user in accordance with the user's use of any of the plurality of Web services.

The information processing device 10 obtains various information from the user device 11, and based on the information, causes the learning model to learn in order to predict a predetermined inclined driver, and based on the learning model, predicts whether any user is a driver with the predetermined inclined driver. In the present embodiment, the information processing device 10 uses the information obtained from the user devices 11-1 to 11-N (i.e., users 1 to N) to cause the learning model to learn. The information processing device 10 is also configured to use the learning model to predict whether the user of the user device 11-T1 (target user T1) is a driver with the predetermined inclined driver. Furthermore, in this embodiment, the source of information required for the learning model (driver tendency prediction model 112 described later) to learn is users 1 to N (user devices 11-1 to 11-N), also referred to as a plurality of sample users.

[Functional structure of information processing device 10] FIG. 2 is a diagram showing an example of the functional structure of the information processing device 10 according to the present embodiment. The information processing device 10 can function as a driver tendency prediction device and/or a learning device. The information processing device 10 shown in FIG. 2 includes: a user feature acquisition unit 101, a vehicle-related feature acquisition unit 102, a level data acquisition unit 103, a learning unit 104, a prediction unit 105, an output unit 106, a learning model storage unit 110, and a feature storage unit 120. The learning model storage unit 110 stores a user feature prediction model 111 and a driver tendency prediction model 112. The various learning models will be described later. In addition, the feature storage unit 120 stores user features 121, vehicle-related features 122, and level data (level features) 123. The various features will be described later.

The user characteristic acquisition unit 101 acquires the user characteristics of the user of the user device 11. In the present embodiment, the so-called user characteristics are factual characteristics (factual information) about the user device or the user, and are characteristics (information) based on facts that are actually or objectively obtained from the user device or the user. The user characteristic acquisition unit 101 can, for example, directly acquire the user characteristics from the user device 11. Furthermore, the user characteristic acquisition unit 101 can acquire the user characteristics as information that the user of the user device 11 is logged into a predetermined Web service. The user characteristics are bound to an identifier of each user (hereinafter referred to as a user ID).

In the present embodiment, the user characteristics acquisition unit 101 acquires demographic information registered in one or more Web services (one or more first services) and information indicating actions regarding service utilization in the one or more Web services (e.g., purchase action information in past Web services) as user characteristics. The one or more Web services from which the user information is acquired may not include vehicle services. Demographic information is information indicating user attributes in terms of demographics, such as gender, age, residential area, occupation, and family composition. Purchase action information is, for example, information regarding purchased items (item name, type, seller, etc.) or information on the date and time of purchase. Furthermore, the user characteristic acquisition unit 101 may also acquire data indicating the user's preferences registered in the Web service by the user at the time of purchase as the user characteristic.

Furthermore, the user feature acquisition unit 101 is configured to apply the acquired user features to the learned user feature prediction model 111, and also acquire the user features (estimated user features (attributes)) estimated for the user features. The user feature prediction model 111 is configured to take the user features (i.e., actual features) of an arbitrary user (referred to as "the first user" for the purpose of explanation) as input, and output the probability (matching probability) that a plurality of user features are respectively consistent (suitable) to the first user. The user feature acquisition unit 101 will eventually determine the estimated user features of the first user based on the matching probability.

For example, the user feature acquisition unit 101 inputs the demographic information of the first user and the purchase history or interest data in the past web service into the user feature prediction model 111 as the user feature of the first user. The user feature prediction model 111 outputs the probability of a plurality of items that the first user is estimated to purchase or a plurality of interests that the first user may have as the probability of matching. Then, the user feature acquisition unit 101 may acquire the items or interests that have a probability greater than a predetermined value as the estimated user feature of the first user. Alternatively, the probability itself may be regarded as the estimated user feature.

The estimated user characteristics according to this embodiment may include characteristics about the user's life stage, such as the probability of working or having children. Here, the probability of having children may be the probability of having children aged 0 to 1, the probability of having children aged 2 to 6, the probability of having children older than elementary school students, and other probabilities classified according to the age of the children. Furthermore, the estimated user characteristics according to this embodiment may include characteristics about the user's lifestyle, such as the probability of traveling, the probability of exercising, the probability of smoking, the probability of drinking, the probability of subscribing to newspapers, and the probability of purchasing health products. Here, the probability of traveling can be classified according to the purpose or type of travel, such as the probability of personal travel, the probability of domestic travel, and the probability of international travel. In addition, the probability of smoking or drinking can be classified according to the location (for example, at home or outside (restaurant, etc.)).

The user feature acquisition unit 101 causes the acquired user features and the estimated user features to be stored in the feature storage unit 120 as user features 121. The user features are stored in association with the user ID. The user feature acquisition unit 101 may also be configured to store the user features 121 in an external device.

The vehicle-related feature acquisition unit 102 acquires information of a vehicle service (second service) registered in one of the Web services by a user of the user device 11 as a vehicle-related feature (vehicle feature). In this embodiment, the vehicle service is a service that provides services related to cars, and users can enjoy the service by registering the vehicle-related features. The vehicle-related features include features such as the manufacturer name, car name, model, contracted automobile insurance company (current and/or past), replacement frequency, and the period from registration to vehicle service (time information). Furthermore, in this embodiment, it includes: the All Cars Service (My Car Service) which is a sub-service included in the vehicle service; the vehicle-related characteristics may include: information on whether the user has logged in to the All Cars Service (1 (logged in) or 0 (not logged in)). In this embodiment, the All Cars Service is a service that gives free points or coupons to users who have registered information about the cars they own (such as the manufacturer name, car name, and model). Users can get free points or coupons by registering the All Cars Service and registering the characteristics of the cars they own. In addition, users who have already registered the vehicle service do not necessarily have to register the All Cars Service.

The vehicle-related feature is bound to the user ID, and the vehicle-related feature acquisition unit 102 causes the acquired vehicle-related feature to be bound to the user ID and stored in the feature storage unit 120 as the vehicle-related feature 122. The vehicle-related feature acquisition unit 102 may also be configured to store the vehicle-related feature 122 in an external device.

The rating data acquisition unit 103 acquires the rating assigned to the user of the user device 11 in the automobile insurance service of one of the Web services as rating data within a legal and appropriate scope. The rating refers to the division required to determine the amount of insurance premium payment. In the case of non-fleet ratings that individuals can join, the rating is divided into levels 1 to 20 in one implementation form. Users who are judged to be safe drivers will be assigned a higher rating. Whether the driving is safe is generally determined based on whether the car owned by the user has been involved in an accident. A user is assigned a predetermined rank when he starts using the service, and is assigned a higher rank after a certain period of time by satisfying predetermined conditions such as no incidents. The rank data acquisition unit 103 is configured to acquire rank data each time the rank is changed (updated).

The level data is bound to the user ID, and the level data acquisition unit 103 causes the acquired level data to be bound to the user ID and stored in the feature storage unit 120 as level data 123. The level data acquisition unit 103 may also be configured to store the level data 123 in an external device.

The learning unit 104 generates learning data (training data) based on the user characteristics 121, the vehicle-related characteristics 122, and the level data 123. Furthermore, the learning unit 104 uses the generated learning data to make the driver tendency prediction model 112 learn. The learning process of the driver tendency prediction model 112 including the generation of the learning data will be described later.

The prediction unit 105 uses the driver inclination prediction model 112 learned by the learning unit 104 to predict whether an arbitrary user is a driver with a predetermined inclination. In this embodiment, the prediction unit 105 predicts whether the target user T1 is a driver with a predetermined inclination. The prediction procedure of whether the target user T1 is a driver with a predetermined inclination will be described later.

The output unit 106 is configured to output the prediction result made by the prediction unit 105. The output can be any output processing, or can be output to an external device through the communication I/F (communication I/F57 in Figure 5), or can be displayed on the display unit (display unit 56 in Figure 5). The output unit 106 can also generate and output information about the prediction result. For example, if the target user T1 is predicted to be a driver with a predetermined tendency, and the target user T1 has not yet joined the car insurance service, the output unit 106 can also create and provide an advertisement representing the car insurance service to the target user T1.

[Learning procedure of driver inclination prediction model] The learning procedure of the driver inclination prediction model 112 performed by the learning unit 104 of this embodiment is explained with reference to FIG. 3A and FIG. 3B . FIG. 3A is a flowchart illustrating the learning procedure of the driver inclination prediction model 112, and FIG. 3B is a conceptual diagram illustrating the learning of the driver inclination prediction model 112. The processing of FIG. 3A can be started after the user feature acquisition unit 101, the vehicle-related feature acquisition unit 102, and the level data acquisition unit 103 respectively store the user feature 121, the vehicle-related feature 122, and the level data 123 in the feature storage unit 120. User characteristics 121, vehicle-related characteristics 122, and rating data 123 respectively include: user characteristics, vehicle-related characteristics, and rating data of a plurality of sample users (users 1 to N).

In FIG. 3A , first, the learning unit 104 acquires features of a plurality of sample users (users 1 to N) required for learning the driver preference prediction model 112 (S31 to S33). In S31, the learning unit 104 acquires user features 121 of a plurality of sample users stored in the feature memory unit 120. In S32, the vehicle-related feature acquisition unit 102 acquires vehicle-related features 122 of a plurality of sample users stored in the feature memory unit 120. In S33, the level data acquisition unit 103 acquires the level data 123 of the plurality of sample users stored in the feature storage unit 120. In addition, the processing of S31 to S33 is performed based on the user ID of each of the plurality of sample users. Furthermore, the processing of S31 to S33 is not limited to the sequence shown in FIG. 3A, and may be performed in a different sequence or simultaneously. Furthermore, in the present embodiment, it is assumed that the features acquired in S31 to S33 have been pre-stored in the feature storage unit 120, but it is not limited thereto. For example, the learning unit 104 may also be configured to directly acquire the features from at least one of the plurality of sample users, or to acquire the features from an external device.

In S34, the learning unit 104 generates (sets) a correct answer label (correct answer data) for the driver tendency prediction model 112 based on the level shown in the level data obtained in S33. In the present embodiment, the predetermined tendency to be predicted is set as a good tendency, and the user with a user ID having a level above the predetermined level is defined as a driver with a good tendency. Therefore, the learning unit 104 sets the probability that the user with a user ID having a level above the predetermined level is a driver with a good tendency to "1" (i.e., the maximum value of the probability) and generates the correct answer label (correct answer data). That is, the learning unit 104 generates a correct answer label of "1" based on the level above the predetermined level. On the other hand, the learning unit 104 generates a correct answer label by setting the probability that the user with the user ID having a level below the predetermined level is a driver with good tendency to "0" (i.e., the minimum value of the probability). That is, the learning unit 104 generates a correct answer label of "0" based on the level below the predetermined level.

The predetermined level is set to, for example, 20 levels. Therefore, a user of level 20 is considered to be a driver with a good tendency (=1), and a user of levels 1 to 19 is considered to be a driver with a bad tendency (=0), and a correct answer label is generated. In addition, the predetermined level can also be input by the operator through the input unit (input unit 55 of FIG. 5) or through the communication unit (communication I/F57 of FIG. 5). Alternatively, the predetermined level is set in advance in the system, and can also be set by an arbitrary program stored in the memory unit (ROM52 or RAM53 of FIG. 5).

In the processing up to S34, user characteristics 121, vehicle-related characteristics 122, and correct answer labels are generated for each of the plurality of sample users (i.e., for each user ID). In the subsequent S35, the learning unit 104 generates data in which the user characteristics 121 and the vehicle-related characteristics 122 are labeled with the generated correct answer labels as learning data. Here, referring to FIG. 3B , the input data 301 contains the user characteristics 121 and the vehicle-related characteristics 122, and the correct answer label 302 corresponds to the generated correct answer label. The learning unit 104 can store the generated learning data in the memory unit (ROM52 or RAM53 in FIG. 5 ).

In S36, the learning unit 104 uses the learning data generated in S35 to make the driver preference prediction model 112 learn. Specifically, referring to FIG. 3B, the learning unit 104 inputs input data 301 containing user features 121 and vehicle-related features 122 for each of a plurality of sample users into the driver preference prediction model 112, and makes the driver preference prediction model 112 learn in such a way that the driver preference prediction model 112 outputs a correct answer label 302.

The driver tendency prediction model 112 is a learning model required for machine learning, and is, for example, a learning model based on CatBoost. Alternatively, the driver tendency prediction model 112 may also be a learning model based on other boosting methods such as XGBoost or LightGBM. By using a learning model based on a boosting method, user characteristics of drivers with good tendencies may also be identified.

The learning unit 104 uses the learning data (input data 301 and correct answer labels 302) of a plurality of sample users to learn the driver tendency prediction model 112. Through learning, the driver tendency prediction model 112 is configured to take the user characteristics 121 and vehicle-related characteristics 122 of an arbitrary user as inputs and output the probability that the arbitrary user is a driver with good tendencies. Once the learning is completed, in S37, the learning unit 104 causes the learned driver tendency prediction model 112 to be stored in the learning model storage unit 110.

[Prediction procedure for a driver with a predetermined inclination] After the driver inclination prediction model 112 is learned by the learning unit 104, the prediction unit 105 uses the learned driver inclination prediction model 112 to predict whether the target user T1 (an arbitrary user) is a driver with a predetermined inclination. In this embodiment, it is predicted whether the target user T1 is a driver with a good inclination. The prediction procedure for a driver with a good inclination performed by the prediction unit 105 of this embodiment is described with reference to FIG. 4A and FIG. 4B. FIG. 4A is a flowchart illustrating a prediction procedure for a driver with a predetermined inclination, and FIG. 4B is a conceptual diagram illustrating a driver inclination prediction model 112 when predicting a predetermined driver.

The process of FIG. 4A is started after the user feature acquisition unit 101 and the vehicle-related feature acquisition unit 102 respectively store the user feature 121 and the vehicle-related feature 122 of the target user T1 in the feature storage unit 120 .

In FIG. 4A , first, the prediction unit 105 obtains the input data 401 for the driver tendency prediction model 112 (S41-S42). In S41, the prediction unit 105 obtains the user feature 121 of the target user T1 stored in the feature memory unit 120. In S42, the vehicle-related feature acquisition unit 102 obtains the vehicle-related feature 122 of the target user T1 stored in the feature memory unit 120. In addition, the processing of S41-S42 is performed based on the user ID of the target user T1. In addition, the processing of S41-S42 is not limited to the order shown in FIG. 4A , and can be performed in a different order or simultaneously. Furthermore, in this embodiment, it is assumed that the features obtained in S41-S42 have been pre-stored in the feature storage unit 120, but this is not limited to this. For example, the prediction unit 105 can also be configured to directly obtain the features from the target user T1, or to obtain the features from an external device.

In S43, the prediction unit 105 inputs the user characteristics 121 and the vehicle-related characteristics 122 of the target user T1 obtained in S41-S42 to the driver tendency prediction model 112, and predicts whether the target user T1 is a driver with a predetermined tendency (a good tendency in the present embodiment). The driver tendency prediction model 112 to which the user characteristics 121 and the vehicle-related characteristics 122 of the target user T1 are input is configured to output the probability that the target user T1 is a driver with a predetermined tendency (the probability 402 of being a driver with a predetermined tendency).

In S44, the prediction unit 105 transmits the prediction result of whether the target user T1 is a driver with a predetermined inclination (i.e., a good inclination) to the output unit 106. For example, the prediction unit 105 may output the probability that the target user T1 is a driver with a good inclination as the prediction result to the output unit 106. Alternatively, a predetermined threshold may be set, and when the probability is greater than the predetermined threshold, the target user T1 is a driver with a good inclination, and when the probability is less than the predetermined threshold, the target user T1 is a driver with a non-good inclination, and the prediction result indicating this is transmitted to the output unit 106. In addition, the predetermined threshold value may be inputted by the operator through the input unit (input unit 55 in FIG. 5 ) or through the communication unit (communication I/F 57 in FIG. 5 ). Alternatively, the predetermined threshold value may be set in advance in the system or may be set by an arbitrary program stored in the memory unit (ROM 52 or RAM 53 in FIG. 5 ).

The output unit 106 to which the prediction result is transmitted in S44 can generate information about the prediction result and output it to an external device such as the user device 11-T1. For example, if the prediction result indicates that the target user T1 is a driver with a favorable tendency and the target user T1 has not yet joined the automobile insurance service, the output unit 106 can also create and provide an advertisement indicating the automobile insurance service to the target user T1. In addition, the output unit 106 can calculate the probability of the target user T1 being a driver with a favorable tendency based on the prediction result for any multiple users including the target user T1, and use the probability to create a new policy in the automobile insurance service.

[Hardware structure of information processing device 10] FIG. 5 is a block diagram of an example of the hardware structure of the information processing device 10 according to the present embodiment. The information processing device 10 according to the present embodiment can also be implemented on a single or multiple computers, mobile devices, or other arbitrary processing platforms. Referring to FIG. 5, although the information processing device 10 is illustrated as an example of being implemented on a single computer, the information processing device 10 according to the present embodiment can also be implemented in a computer system containing multiple computers. Multiple computers can be connected via a wired or wireless network to communicate with each other.

As shown in FIG5 , the information processing device 10 may include: CPU 51, ROM 52, RAM 53, HDD 54, input unit 55, display unit 56, communication I/F 57, and system bus 58. The information processing device 10 may also include an external memory. CPU (Central Processing Unit) 51 is the central processing unit that controls the operation of the information processing device 10 and controls each component (52~57) through the data transmission path, that is, the system bus 58.

ROM (Read Only Memory) 52 is a non-volatile memory that stores control programs and the like necessary for CPU 51 to execute processing. In addition, the program can also be stored in a non-volatile memory such as HDD (Hard Disk Drive) 54, SSD (Solid State Drive) or an external memory such as a removable storage medium (not shown). RAM (Random Access Memory) 53 is a volatile memory that functions as the main memory, work area, etc. of CPU 51. That is, when CPU 51 is executing processing, it loads necessary programs and the like from ROM 52 into RAM 53, and various functional actions are realized by executing the programs and the like. The learning model memory unit 110 and the feature memory unit 120 shown in FIG2 can be constructed by RAM53.

HDD54 stores various data or information necessary for, for example, CPU51 to process using a program. HDD54 also stores various data or information obtained by, for example, CPU51 processing using a program. Input unit 55 is composed of a pointing device such as a keyboard or a mouse. Display unit 56 is composed of a screen such as a liquid crystal display (LCD). Display unit 56 can also be combined with input unit 55 to function as a GUI (Graphical User Interface).

The communication I/F57 is an interface for controlling the communication between the information processing device 10 and the external device. The communication I/F57 provides an interface with the network and performs communication with the external device through the network. Various data or various parameters are sent and received with the external device through the communication I/F57. In this embodiment, the communication I/F57 can perform communication through a wired LAN (Local Area Network) or a dedicated line based on the communication specifications of Ethernet (registered trademark) and the like. However, the network that can be used in this embodiment is not limited to this, and can also be composed of a wireless network. The wireless network includes wireless PAN (Personal Area Network) such as Bluetooth (registered trademark), ZigBee (registered trademark), and UWB (Ultra Wide Band). It also includes wireless LAN (Local Area Network) such as Wi-Fi (Wireless Fidelity) (registered trademark), or wireless MAN (Metropolitan Area Network) such as WiMAX (registered trademark). It even includes wireless WAN (Wide Area Network) such as LTE/3G, 4G, and 5G. In addition, as long as the network can connect and communicate with each other, the specifications, scale, and structure of the communication are not limited to the above.

At least a part of the functions of each element of the information processing device 10 shown in FIG2 can be realized by executing a program by the CPU 51. However, at least a part of the functions of each element of the information processing device 10 shown in FIG2 can also be operated as dedicated hardware. In this case, the dedicated hardware is operated based on the control of the CPU 51.

[Hardware structure of user device 11] The hardware structure of user device 11 shown in FIG1 may be the same as that of FIG5. That is, user device 11 may include: CPU 51, ROM 52, RAM 53, HDD 54, input unit 55, display unit 56, communication I/F 57, and system bus 58. User device 11 may display various information provided by information processing device 10 on display unit 56, and perform processing corresponding to input operations received from users through GUI (structure of input unit 55 and display unit 56).

In this way, the information processing device 10 can predict whether the target user is a driver with a predetermined tendency based on the user characteristics of the target user obtained through the Web service and the information related to the car owned by the target user (car-related characteristics). In this way, even if there is no objective data such as physical data obtained through driving or the presence or absence of accidents, it is still possible to predict whether the target user is a driver with a predetermined tendency.

Furthermore, the information processing device 10 can provide various new services to users based on the prediction results of drivers with a given tendency. For example, if the information processing device 10 predicts that a target user is a driver with a good tendency and the target user has not yet joined a specific automobile insurance service, the information processing device 10 can recommend the specific automobile insurance service to the target user. Furthermore, the information processing device 10 can calculate the probability of a driver with a good tendency for any plurality of users including the target user based on the prediction results for the plurality of users, and use the probability to create a new policy in the automobile insurance service. This can improve the profitability of the automobile insurance service.

Furthermore, in the present embodiment, a learning model required for machine learning is constructed for the prediction, that is, a driver preference prediction model 112. Here, the driver preference prediction model 112 is learned using learning data generated from user characteristics, vehicle-related characteristics, and grades in automobile insurance services obtained from a plurality of users. The learning data contains correct answer labels created based on grades actually assigned in automobile insurance services, so that the driver preference prediction model 112 can be learned in a manner that takes into account the context in which the grades are assigned and outputs a more accurate prediction result.

In addition, although the above embodiments are described in relation to automobiles, the above embodiments are applicable to any vehicle. For example, the above embodiments are applicable to other vehicles such as bicycles and motorcycles.

In the above-mentioned embodiment, the driver tendency prediction model 112 is configured to output the probability that the target user is a driver with a predetermined tendency by taking the user characteristics 121 and the vehicle-related characteristics 122 of the target user as inputs, as shown in FIG. 3B or FIG. 4B , but the input data is not limited thereto. For example, the model may be configured and learned to output the probability that the target user is a driver with a predetermined tendency by taking only the user characteristics 121 of the target user as inputs.

In addition, although a specific implementation form is described above, the implementation form is merely an example and is not intended to limit the scope of the present invention. The devices and methods described in this specification may also be embodied in forms other than those described above. Furthermore, the implementation forms described above may be appropriately omitted, replaced, and modified without departing from the scope of the present invention. The forms in which the omissions, replacements, and modifications are made are included in the scope of the matters described in the patent application and their equivalents, and still fall within the technical scope of the present invention.

The present embodiment discloses the following configuration. [1] A driver tendency prediction device comprises: user characteristic acquisition means for acquiring user characteristics representing characteristics of a target user from one or more first services; and vehicle characteristic acquisition means for acquiring vehicle characteristics representing characteristics of a vehicle owned by the target user from a second service different from the one or more first services; and prediction means for predicting whether the target user is a driver with the predetermined tendency by using a learning model that takes the target user characteristics and the vehicle characteristics as inputs and outputs the probability that the target user is a driver with the predetermined tendency after learning.

[2] A driver tendency prediction device as described in [1], wherein the previous user characteristics include: factual characteristics of the previous target user and inferred characteristics inferred based on the previous factual characteristics.

[3] A driver tendency prediction device as described in [2], wherein the aforementioned factual features include: demographic information of the aforementioned target user who is logged in to one or more aforementioned first services, and information indicating actions regarding service utilization in the aforementioned first services.

[4] A driver tendency prediction device as described in [2] or [3], wherein the presumed characteristics include at least one of characteristics related to the life stage of the user of the presumed object and characteristics related to the lifestyle of the user of the presumed object.

[5] A driver tendency prediction device as described in any one of [1] to [4], wherein a plurality of services including one or more of the aforementioned first services and the aforementioned second service are configured to manage information using a user identifier that is common among the plurality of services.

[6] A learning device comprising: user feature acquisition means for acquiring user features representing features of a plurality of users from one or more first services; and vehicle feature acquisition means for acquiring vehicle features representing features of vehicles owned by the plurality of users from a second service different from the one or more first services. ; and level acquisition means for acquiring vehicle insurance levels of vehicles owned by a plurality of users; and generation means for generating learning data based on the previous user characteristics, the previous vehicle characteristics, and the previous levels; and learning means for using the previous learning data to learn a learning model required for predicting the probability of a driver with a given tendency.

[7] A learning device as described in [6], wherein the previous user characteristics include: factual characteristics about the previous plurality of users, and inferred characteristics inferred based on the previous factual characteristics.

[8] A learning device as described in [7], wherein the aforementioned factual features include: demographic information of the aforementioned target user who is logged in to one or more aforementioned first services, and information indicating actions related to service utilization in the aforementioned first services.

[9] A learning device as described in [7] or [8], wherein the aforementioned presumed features include at least one of features related to the life stages of the aforementioned plurality of users and features related to the lifestyles of the aforementioned plurality of users.

[10] A learning device as described in any one of [6] to [9], wherein a plurality of services including one or more of the first service and the second service are configured to manage information using a user identifier that is common among the plurality of services.

1~N: User T1: Target user 10: Information processing device 11, 11-1~11-N, 11-T1: User device 51: CPU 52: ROM 53: RAM 54: HDD 55: Input unit 56: Display unit 57: Communication I/F 58: System bus 101: User feature acquisition unit 102: Vehicle-related feature acquisition unit 103: Level data acquisition unit 104: Learning unit 105: Prediction unit 106: Output unit 110: Learning model storage unit 111: User feature prediction model 112: Driver tendency prediction model 120: Feature storage unit 121: User characteristics 122: Vehicle-related characteristics 123: Level data 301: Input data 302: Correct answer label 401: Input data 402: Probability of being a driver with a given tendency

[FIG. 1] FIG. 1 is a diagram illustrating an example of a configuration of an information processing system according to an embodiment. [FIG. 2] FIG. 2 is a diagram illustrating an example of a functional configuration of an information processing device according to an embodiment. [FIG. 3A] FIG. 3A is a flowchart illustrating a learning procedure of a driver inclination prediction model. [FIG. 3B] FIG. 3B is a conceptual diagram illustrating the learning of a driver inclination prediction model. [FIG. 4A] FIG. 4A is a flowchart illustrating a prediction procedure of a driver with a predetermined inclination. [FIG. 4B] FIG. 4B is a conceptual diagram illustrating a driver inclination prediction model when predicting a driver with a predetermined inclination. [Figure 5] Figure 5 is a diagram showing an example of the hardware configuration of an information processing device and a user device according to an embodiment.

112: Driver tendency prediction model

121: User characteristics

122: Car-related features

401: Input data

402: Probability of being a driver of a given tendency

T1: Target user

Claims (14)

A driver tendency prediction device is characterized by comprising: user characteristic acquisition means for acquiring user characteristics representing characteristics of a target user from one or more first services; and vehicle characteristic acquisition means for acquiring vehicle characteristics representing characteristics of a vehicle owned by a previous target user from a second service different from the one or more first services; and prediction means for predicting whether the previous target user is a driver with a previously specified tendency by using a learning model that takes the previous user characteristics and the previous vehicle characteristics as inputs and outputs the probability that the previous target user is a driver with a specified tendency after learning. As described in claim 1, the driver tendency prediction device, wherein the previous user characteristics include: factual characteristics of the previous target user, and inferred characteristics inferred based on the previous factual characteristics. The driver tendency prediction device as described in claim 2, wherein the aforementioned factual features include: demographic information of the aforementioned target user registered in one or more aforementioned first services, and information indicating actions regarding service utilization in the aforementioned first services. The driver tendency prediction device as recited in claim 2 or 3, wherein the presumed characteristics include at least one of the characteristics of the life stage of the user of the presumed object and the characteristics of the lifestyle of the user of the presumed object. The driver tendency prediction device as described in claim 1, wherein a plurality of services including the aforementioned one or more first services and the aforementioned second service are configured so that information can be managed using a user identifier that is common among the plurality of services. A learning device, characterized by comprising: user feature acquisition means for acquiring user features representing features of a plurality of users from one or more first services; and vehicle feature acquisition means for acquiring vehicle features representing features of vehicles owned by the plurality of users from a second service different from the one or more first services. and a level acquisition means for acquiring the levels of vehicle insurance of vehicles owned by a plurality of users; and a generation means for generating learning data based on the aforementioned user characteristics, the aforementioned vehicle characteristics, and the aforementioned levels; and a learning means for using the aforementioned learning data to enable a learning model required for predicting the probability of a driver with a given tendency to learn. A learning device as recited in claim 6, wherein the aforementioned user characteristics include: factual characteristics about the aforementioned plurality of users, and inferred characteristics inferred based on the aforementioned factual characteristics. A learning device as recited in claim 7, wherein the aforementioned factual features include: demographic information of the aforementioned plurality of users logged in to the aforementioned one or more first services, and information indicating actions regarding service utilization in the aforementioned first services. A learning device as recited in claim 7 or 8, wherein the aforementioned presumed features include at least one of features related to the life stages of the aforementioned plurality of users and features related to the lifestyles of the aforementioned plurality of users. A learning device as recited in claim 6, wherein a plurality of services including one or more of the aforementioned first services and the aforementioned second services are configured so that information can be managed using a user identifier that is common among the plurality of services. A driver tendency prediction method is a driver tendency prediction method executed by an information processing device, characterized by comprising: a user characteristic acquisition step of acquiring user characteristics representing characteristics of a target user from one or more first services; and a vehicle characteristic acquisition step of acquiring characteristics representing characteristics of a vehicle owned by the target user. The vehicle characteristics to be collected are obtained from a second service that is different from one or more first services recorded previously; and the prediction step is to predict whether the target user recorded previously is a driver with the specified inclination by using a learning model that takes the previous user characteristics and the previous vehicle characteristics as inputs and outputs the probability that the target user recorded previously is a driver with the specified inclination after learning. A learning method is a learning method executed by an information processing device, characterized by comprising: a user feature acquisition step of acquiring user features representing features of a plurality of users from one or more first services; and a vehicle feature acquisition step of acquiring vehicle features representing features of vehicles owned by the plurality of users from one or more first services. The second service of the service is obtained; and the level acquisition step is to obtain the vehicle insurance levels of the vehicles owned by the plurality of users; and the generation step is to generate learning data based on the user characteristics, the vehicle characteristics, and the levels; and the learning step is to use the learning data to learn the learning model required for predicting the probability of a driver with a given tendency. An information processing program is an information processing program required for causing a computer to execute information processing, and the program is used to cause the aforementioned computer to execute the following processing: a user characteristic acquisition processing is to acquire user characteristics representing characteristics of a target user from one or more first services; and a vehicle characteristic acquisition processing is to acquire information representing characteristics of a vehicle owned by the aforementioned target user. The vehicle characteristics of the vehicle are obtained from a second service that is different from one or more first services mentioned previously; and the prediction processing is to use a learning model that takes the previous user characteristics and the previous vehicle characteristics as inputs and outputs the probability that the previous target user is a driver with a predetermined tendency after learning, to predict whether the previous target user is a driver with a predetermined tendency. An information processing program is an information processing program required for causing a computer to execute information processing, and the program is used to cause the aforementioned computer to execute the following processing: user characteristic acquisition processing is to acquire user characteristics representing characteristics of a plurality of users from one or more first services; and vehicle characteristic acquisition processing is to acquire vehicle characteristics representing characteristics of vehicles owned by the aforementioned plurality of users from one or more first services. The second service of the first service or more is obtained; and the level acquisition process is to obtain the vehicle insurance levels of the vehicles owned by the plurality of users; and the generation process is to generate learning data based on the user characteristics, the vehicle characteristics, and the levels; and the learning process is to use the learning data to learn the learning model required for predicting the probability of a driver with a given tendency.

Images