Automobile recommendation system

Abstract

An automobile recommendation system (1) recommends an automobile to a user and is provided with: a memory for storing an instruction; and one or more processors. When the instruction is executed, the one or more processors cause the automobile recommendation system (1) to: determine first-group users belonging to an attribute that is the same as or similar to an attribute of a specific user at a prescribed time point; calculate an indicator that represents preference of the user from the difference between a characteristic value of a vehicle type requested by the first-group users and a characteristic value of a vehicle type requested by the specific user at the prescribed time point; determine second-group users belonging to an attribute that is the same as or similar to the current attribute of the specific user; calculate an added characteristic value of the vehicle type on the basis of a characteristic value of a vehicle type requested by the second-group users and the indicator that represents the preference of the specific user; determine a candidate vehicle type on the basis of the added characteristic value of the vehicle type and a characteristic value of each vehicle type; and recommend the candidate vehicle type to the specific user.

Description

Car recommendation system

[0001] The present invention relates to a technology for recommending products to individual users, and in particular to a technology that is effective when applied to an automobile recommendation system that recommends automobiles.

[0002] When selling products, systems are being considered that recommend appropriate products based on the user's needs and values. However, user needs and values ​​cannot usually be directly observed unless asked through a questionnaire or other means.

[0003] In response to this, for example, Japanese Patent Application Laid-Open No. 2022-90717 (Patent Document 1) describes a method of estimating the characteristics (needs) desired by a user for a specific characteristic item of a product (e.g., an automobile) based on the user's attribute information, and selectively suggesting products that meet the user's needs by comparing the estimated needs with the actual characteristics of each product.

[0004] Japanese Patent Application Laid-Open No. 2022-90717

[0005] According to the conventional technology, it is possible to estimate the needs of a user based on the attributes of the user and recommend products that meet the needs to the user.

[0006] In this case, to estimate needs (tendencies) from user attributes, for example, a group of users with the same or similar attributes is segmented and the individual needs of these users are averaged. However, it is not necessarily the case that the same products should be recommended to users with the same or similar attributes. Even if the attributes are the same, individual users have different values ​​(tastes), so it is necessary to recommend products that match those values. However, such values ​​cannot usually be observed directly unless the user is asked.

[0007] Therefore, an object of the present invention is to provide an automobile recommendation system that recommends products (especially automobiles) by inferring differences and changes in needs due to differences in user attributes and the values ​​of individual users.

[0008] The above and other objects and novel features of the present invention will become apparent from the description of this specification and the accompanying drawings.

[0009] Among the inventions disclosed in this application, the outline of representative inventions will be briefly explained as follows.

[0010] A representative embodiment of the present invention is an automobile recommendation system that includes a memory for storing instructions and one or more processors. When the instructions are executed, the one or more processors cause the automobile recommendation system to perform the following tasks: determine a first group of users who belong to attributes that are the same as or similar to the attributes of a specific user at a given time; calculate an index indicating the user's values ​​from the difference between the characteristic values ​​of the vehicle models requested by the users of the first group and the characteristic values ​​of the vehicle models requested by the specific user at the given time; determine a second group of users who belong to attributes that are the same as or similar to the current attributes of the specific user; calculate an aggregated characteristic value of the vehicle models based on the characteristic values ​​of the vehicle models requested by the users of the second group and the index indicating the specific user's values; determine candidate vehicle models based on the aggregated characteristic value of the vehicle models and the characteristic values ​​of each vehicle model; and recommend the candidate vehicle models to the specific user.

[0011] The effects obtained by the representative inventions disclosed in this application can be briefly explained as follows.

[0012] That is, according to the representative embodiment of the present invention, it is possible to recommend cars by inferring differences and changes in needs due to differences in user attributes and the values ​​of individual users.

[0013] 1A and 1B are diagrams showing an overview of an example configuration of an automobile recommendation system according to an embodiment of the present invention. (a) and (b) are diagrams showing an overview of an example of the concept of recommending automobiles by inferring user needs and values ​​according to an embodiment of the present invention. (a) is a diagram showing an overview of an example data configuration of performance specifications for each vehicle model according to an embodiment of the present invention. (b) is a diagram showing an overview of an example data configuration of characteristic values ​​for each vehicle model according to an embodiment of the present invention. (b) is a diagram showing an overview of an example data configuration of characteristic values ​​for each vehicle model according to an embodiment of the present invention. (b) is a diagram showing an overview of an example data configuration of characteristic values ​​for each user ... 1 is a flowchart showing a series of operations of a recommendation process executed in an automobile recommendation system according to an embodiment of the present invention.

[0014] Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings. In all drawings used to explain the embodiments, the same parts are generally designated by the same reference numerals, and repeated explanations will be omitted. However, parts that have been designated and explained in one drawing may be referred to by the same reference numerals in the explanation of other drawings, although they will not be shown again.

[0015] <Summary> As mentioned above, it is usually not possible to directly observe the needs of different users or the values ​​of individual users. This is because it is necessary to define what "needs" and individual "values" are, and even if these can be defined, it is usually not possible to directly observe them without asking the users directly.

[0016] In contrast, the automobile recommendation system according to one embodiment of the present invention, when recommending automobiles to purchase to a user, considers that the user's needs and values ​​are reflected in the performance indexes of the vehicle model owned by the user. In other words, since it is believed that a user will own an automobile that matches their own values, it is believed that the user's needs and values ​​can be discovered based on the performance indexes of the vehicle model currently owned.

[0017] The needs due to differences in user attributes can be found as representative values ​​(such as averages or medians) of automobile performance indicators required by a group of users with the same or similar attributes (hereinafter sometimes simply referred to as "same attributes"), and the values ​​of each individual user can be evaluated as the difference between the performance indicators required by that user and the representative values ​​(for example, "this user places more importance on fuel efficiency than the average user in their 30s").The values ​​of this user are considered to be derived from something that the user is innately endowed with, and will not change significantly even if the user's life stage changes.

[0018] FIG. 2 is a diagram outlining an example of the concept of recommending a vehicle by inferring a user's needs and values ​​in one embodiment of the present invention. FIG. 2(a) shows an example of inferring a user's values. The horizontal axis of the graph shows vehicle characteristics (four in the example of FIG. 2: "Characteristic 1" to "Characteristic 4"), and the vertical axis shows the strength of each characteristic. For each characteristic, the open graph shows the vehicle characteristics (representative value) desired by a group of users with the same attributes as the target user at the time the target user started owning the vehicle they currently own (at the time of ownership start). The shaded graph shows the vehicle characteristics desired by the individual user at the time of ownership start.

[0019] In this embodiment, the difference between the open graph and the shaded graph for each characteristic item in this graph, i.e., the difference between the representative value (tendency) of a group of users with the same attribute and the characteristic value desired by the individual user, is considered to represent the user's values. In the example of Figure 2(a), there is a difference between "Characteristic 1" and "Characteristic 3" (the black arrow in the figure), and these represent the user's values. Note that the reason why the time when the user first owned the car they currently own is used as the basis here is because at the time of ownership, there is an opportunity to obtain various information about the user's attributes through sales activities, various procedures, etc., and therefore the time when ownership began is not necessarily the basis, and other points in time can also be used as the basis.

[0020] 2(b) shows an example of determining the characteristics of a vehicle model currently recommended to the user based on the user's values ​​obtained in FIG. 2(a). The white graph shows the vehicle model characteristics (representative values) desired by a group of users with the same attributes as the current user. The strength of the characteristics of each characteristic item differs from that of the white graph in FIG. 2(a), but this reflects differences and changes in attributes such as age (for example, needs change as people move from their 20s to their 30s due to changes in life stages, etc.).

[0021] In this embodiment, by reflecting the user's values ​​(black arrows) obtained in Fig. 2A on the open graph (vehicle characteristics desired by a group of users with the same attributes), it is possible to obtain the characteristics of a recommended vehicle (shaded graph) that reflect the user's values. Then, a vehicle that actually has characteristics close to (similar to) the characteristics of the recommended vehicle obtained here becomes the vehicle to be recommended.

[0022] In this way, by separating the tendencies of each user's attributes from the values ​​of each individual user, it becomes possible to recommend cars according to changes in life stages. For example, it is possible to make recommendations that can be interpreted as follows: "As you have changed age, your next car should be a car like this, in line with general user tendencies. However, since you are a user who places importance on environmental performance, we will recommend a car that is relatively fuel-efficient." Or, "If a user who drove a car like this when they lived in Tokyo moves to Hokkaido, we will recommend a car like this," or "If a user who drove a car like this as a single person gets married and starts living with their spouse, we will recommend a car like this," or "For a user who enjoys camping, we will recommend a car model that prioritizes space over a group of users with the same attributes."

[0023] <System Configuration and Processing Flow> Figure 1 is a diagram showing an overview of an example configuration of a car recommendation system according to one embodiment of the present invention. The car recommendation system 1 is an information processing system that is configured, for example, by a server device, a virtual server constructed on a cloud computing service, or an information processing terminal, and that realizes various functions related to car recommendations by executing middleware such as an OS (Operating System), a DBMS (Database Management System), and a Web server program, which are deployed on memory from a recording device such as an HDD (Hard Disk Drive) or an SSD (Solid State Drive), and software running thereon, using a CPU (Central Processing Unit) not shown.

[0024] The automobile recommendation system 1 includes various units, such as a vehicle model characteristic calculation unit 11, a user-requested vehicle model characteristic calculation unit 12, a user attribute-requested vehicle model characteristic calculation unit 13, a user value calculation unit 14, a current user-requested vehicle model characteristic calculation unit 15, and a recommended vehicle model calculation unit 16, which are implemented as software. The system also includes various data stores, such as a vehicle model performance specification 21, a user-owned vehicle model 22, user attributes at the time of ownership start 23, current user attributes 24, vehicle model characteristic values ​​31, user-requested vehicle model characteristics 32, user attribute-requested vehicle model characteristics 33, user value calculation unit 34, current user-requested vehicle model characteristics 35, and a recommended vehicle model 36, which are configured using databases, files, etc.

[0025] In the process of recommending automobiles to users, first, the vehicle model characteristic calculation unit 11 calculates characteristic values ​​for each vehicle model of each manufacturer in advance. Here, data on the performance specifications for each vehicle model held in the vehicle model performance specifications 21 is used as input, and characteristic values ​​for each vehicle model are calculated based on a predetermined model obtained by machine learning and output as vehicle model characteristic values ​​31. There are no particular limitations on the tools, libraries, etc. used for machine learning, and available tools, libraries, etc. may be used as appropriate.

[0026] FIG. 3 is a diagram outlining an example of the data structure of vehicle-specific performance specifications 21 in one embodiment of the present invention. The vehicle-specific performance specifications 21 is a table that holds data on performance indexes and performance specifications for each vehicle model of each manufacturer, and includes, for example, key items such as manufacturer, body shape, vehicle model name, and model year (items marked with * in the figure; the same applies below), as well as items related to performance specifications such as new vehicle price, horsepower, fuel economy, overall length, overall width, and overall height. Note that the performance specification items are not limited to these, and other items may also be included. Data for each of these items can be collected automatically or manually from various media, such as the websites and catalogs of each automobile manufacturer, commercially available data, etc.

[0027] FIG. 4 is a diagram outlining an example of the data configuration of the vehicle-specific characteristic values ​​31 in one embodiment of the present invention. The vehicle-specific characteristic values ​​31 are a table that holds characteristic values ​​calculated for each vehicle model by the vehicle-specific characteristic calculation unit 11 based on the data of the vehicle-specific performance specifications 21. The table includes, for example, key items such as manufacturer, body shape, vehicle model name, and model year, as well as characteristic value items such as characteristic 1, characteristic 2, characteristic 3, and so on. In this embodiment, values ​​for each of the characteristic 1, characteristic 2, characteristic 3, and so on are calculated based on a predetermined model obtained by machine learning, so the meaning of each item and value is a black box and cannot be interpreted by humans. The number of items is not particularly limited and is determined according to the design of the model.

[0028] Returning to FIG. 1 , next, the data of the vehicle model currently owned by each user and the user's attribute information at the time of starting ownership, which are stored in the user-owned vehicle model 22 and the user attributes at the time of starting ownership 23, are input, and the user-requested vehicle model characteristic calculation unit 12 refers to the above-mentioned vehicle model characteristic value 31 to calculate what characteristics the vehicle model owned by each user has, i.e., what vehicle model characteristics each user is requesting, and outputs the result as the user-requested vehicle model characteristics 32.

[0029] 5 is a diagram outlining an example of the data structure of the user-owned vehicle model 22 in one embodiment of the present invention. The user-owned vehicle model 22 is a table that stores information about vehicle models owned by each user, and includes, for example, a key item of a user ID and items that specify the vehicle model, such as the manufacturer of the vehicle, the body shape of the vehicle, the name of the vehicle, and the model year of the vehicle.

[0030] 6 is a diagram outlining an example of the data configuration of the user attributes 23 at the time of ownership start in one embodiment of the present invention. The user attributes at the time of ownership start 23 is a table that holds information on the attributes of each user at the time when that user started owning the vehicle model, and has, for example, a key item of user ID and items related to user attributes such as gender, prefecture at the time of ownership start, and age category at the time of ownership start. The items related to user attributes are not limited to these, and may also include other attribute items such as occupation, family composition, and hobbies.

[0031] 7 is a diagram outlining an example of the data structure of the user-requested vehicle model characteristics 32 in one embodiment of the present invention. The user-requested vehicle model characteristics 32 is a table that stores information on the attributes of each user and the characteristics of the vehicle models owned, i.e., information on what characteristics each user required for the vehicle model at the time they began owning the vehicle model. For example, the table includes a key field for user ID, fields related to user attributes such as gender, prefecture at the time of ownership start, and age category at the time of ownership start, fields specifying the vehicle model such as the owned vehicle manufacturer, the owned vehicle body shape, the owned vehicle name, and the owned vehicle model year, and fields for characteristic values ​​such as characteristic 1, characteristic 2, characteristic 3, etc. Each of the above fields has the same content as the corresponding fields in the user-owned vehicle model 22, user attributes at the time of ownership start 23, and vehicle model characteristic values ​​31, which are input.

[0032] Returning to FIG. 1 , the above-mentioned user-requested vehicle model characteristics 32 are input, and the user attribute-requested vehicle model characteristics calculation unit 13 performs aggregation and statistical processing on a user attribute basis, and outputs the result as user attribute-requested vehicle model characteristics 33.

[0033] 8 is a diagram outlining an example of the data configuration of the required vehicle model characteristics for each user attribute 33 in one embodiment of the present invention. The required vehicle model characteristics for each user attribute 33 is a table that stores data on the characteristics of the vehicle models owned by each user, which have been tabulated and statistically processed for each user attribute at the time of ownership start, i.e., information on needs (tendencies) due to differences in user attributes, and has key items such as gender, prefecture at the time of ownership start, and age category at the time of ownership start, as well as items for characteristic values ​​such as characteristic 1, characteristic 2, characteristic 3, ...

[0034] In this embodiment, the user attributes to be counted are gender, prefecture, and age category (for example, "20s," "30s," "40s," etc.), but are not limited to this. The user's residential area may be divided into smaller areas rather than prefectures, or conversely, into larger areas such as Kanto, Tohoku, Chubu, etc. Other attribute items such as occupation and family structure may also be used. The values ​​of each item, characteristic 1, characteristic 2, characteristic 3, etc., may be counted by user attribute and used as the average value, or other statistical indicators such as the median may be used.

[0035] Returning to FIG. 1 , next, the user-specific desired vehicle model characteristics 32 and the user-specific desired vehicle model characteristics 33 are input, and the user-specific value calculation unit 14 calculates each user's value (more accurately, an "index indicating the value," but will be simply referred to as "value") and outputs it as the user-specific value 34.

[0036] 9 is a diagram outlining an example of the data configuration of the user-specific values ​​34 in one embodiment of the present invention. The user-specific values ​​34 is a table that stores information on the differences between the representative values ​​of vehicle model characteristics desired by a group of users with the same attributes as the user and the vehicle model characteristics requested by the user. For example, the table has a key item of a user ID and characteristic value items such as characteristic 1, characteristic 2, characteristic 3, etc. The values ​​of the characteristic 1, characteristic 2, characteristic 3, etc. items in this table are the differences between the values ​​of the characteristic 1, characteristic 2, characteristic 3, etc. items in the user attribute-specific required vehicle model characteristics 33 (needs (tendencies) based on the attributes of the user at the start of vehicle ownership) and the values ​​of the characteristic 1, characteristic 2, characteristic 3, etc. items in the user-specific required vehicle model characteristics 32 (vehicle model characteristics requested by the user at the start of vehicle ownership), as shown in FIG. 2A.

[0037] Returning to Figure 1, next, using the above-mentioned user-specific values ​​34 and data on current user attributes 24 as input, and referring to the above-mentioned user attribute-specific requested vehicle model characteristics 33, the current user requested vehicle model characteristics calculation unit 15 calculates information on the vehicle model characteristics requested by each current user, i.e., information on the vehicle model characteristics to be recommended to the current user, and outputs it as current user requested vehicle model characteristics 35.

[0038] 10 is a diagram outlining an example of the data configuration of the current user characteristics 24 in one embodiment of the present invention. The current user characteristics 24 is a table that holds current attribute information for each user, and includes, for example, a key item of user ID and items related to user attributes such as gender, current age, and current prefecture. As described above, the attribute items are not limited to those shown in the example of FIG. 10; other attribute items such as occupation and family structure may be used, and residential areas may be identified using divisions other than prefectures.

[0039] 11 is a diagram outlining an example of the data configuration of the current user-requested vehicle model characteristics 35 in one embodiment of the present invention. The current user-requested vehicle model characteristics 35 is a table that holds information on each user's current attributes and the characteristics of the vehicle model requested, i.e., information on the vehicle model characteristics that each user is currently requesting, and has, for example, a key item for user ID, items related to user attributes such as gender, current age, and current prefecture, and items for characteristic values ​​such as characteristic 1, characteristic 2, characteristic 3, and so on.

[0040] 2(b), the values ​​of each item of characteristic 1, characteristic 2, characteristic 3, ... are obtained by adding the values ​​of each item of characteristic 1, characteristic 2, characteristic 3, ... in the user-specific value system 34 of the target user (the user's values) to the values ​​of each item of characteristic 1, characteristic 2, characteristic 3, ... obtained from the user attribute request vehicle model characteristics 33 based on the target user's current attribute information (needs (tendencies) based on the user's current attributes). In other words, the values ​​of each item of characteristic 1, characteristic 2, characteristic 3, ... here indicate the characteristics of the vehicle model to be recommended to the user.

[0041] Returning to FIG. 1 , next, the current user-requested vehicle model characteristics 35 are input, and the recommended vehicle model calculation unit 16 calculates specific vehicle model candidates to be recommended by referring to the vehicle model characteristic values ​​31, and further refers to the vehicle model performance specifications 21 to output a specific recommended vehicle model 36. Here, one or more vehicle models having characteristic values ​​similar to the values ​​of each item of characteristic 1, characteristic 2, characteristic 3, ... in the current user-requested vehicle model characteristics 35 are extracted from the vehicle model characteristic values ​​31 and used as candidates for the specific vehicle model to be recommended. Note that the method for determining similarity is not particularly limited, and an appropriate method can be used, for example, by vectorizing the values ​​of each item of characteristic 1, characteristic 2, characteristic 3, ... as elements and evaluating the similarity of the vectors.

[0042] 12 is a diagram outlining an example of the data configuration of the recommended vehicle model 36 in one embodiment of the present invention. The recommended vehicle model 36 is a table that holds information about vehicle models and their characteristics recommended for each user, and includes key items such as user ID, manufacturer, body shape, vehicle model name, and model year, as well as performance specifications such as new vehicle price, horsepower, fuel efficiency, overall length, overall width, and overall height.

[0043] Note that the examples of data structures shown in Figures 3 to 12 show logical table structures (views), and in reality, for example, multiple logical tables in a database may be implemented in a normalized form as one or more physical tables.

[0044] Furthermore, in this embodiment, the case where each user currently owns a car, i.e., a case where a user is replacing a car, is taken as an example, but this is not limited to this. As long as information that allows the characteristics of the car model desired by the user can be obtained, it is not necessary for the user to currently own a car, and the system can also be applied to the case of purchasing a new car. For example, if it is determined from the user's web browsing history that the user is intensively browsing a specific car model, the characteristics of that car model may be treated as the characteristics of the car model desired by the user.

[0045] 13, an example of the hardware configuration of, for example, a server device included in the car recommendation system 1 will be described. The server device 1300 includes a memory 1301, a processor 1302, a communication interface 1303, a storage 1304, an input interface 1305, a bus 1306, and the like.

[0046] The memory 1301 is a volatile memory such as a dynamic random access memory (DRAM) that temporarily stores programs and calculation results of a processor such as a CPU. The processor 1302 includes one or more processors such as a CPU. The communication interface 1303 includes an interface for wired or wireless communication and communicates with, for example, an information processing terminal via the Internet or a specific network. The communication interface 1303 includes a communication circuit for realizing communication with, for example, an information processing terminal. The storage 1304 includes a recording device such as an HDD or SSD and stores, for example, middleware such as an OS, DBMS, or web server program, or software for recommending automobiles. The input interface 1305 includes a keyboard that accepts operations by an administrator managing the server device, but this may be unnecessary if the server device is operated remotely via a network. The bus 1306 transmits data between blocks of the server device.

[0047] When the processor 1302 expands and executes the computer program stored in the storage 1304 in the memory 1301, the functions of each part, such as the vehicle model characteristics calculation part 11, the user-requested vehicle model characteristics calculation part 12, the vehicle model characteristics calculation part 13 requested by each user attribute, the user-requested value calculation part 14, the current user-requested vehicle model characteristics calculation part 15, and the recommended vehicle model calculation part 16, which were described above with reference to Figure 1, are realized.

[0048] <Series of Operations in Car Recommendation Processing> A series of operations in the car recommendation processing (also simply referred to as recommendation processing) executed in the car recommendation system will be described with reference to Fig. 14. The recommendation processing is realized, for example, by the processor 1302 of the server device 1300 loading a computer program stored in the storage 1304 into the memory 1301 and executing it.

[0049] 14 is started, the processor 1302 calculates characteristic values ​​of individual vehicle models by applying a machine learning model to data of performance specifications of the individual vehicle models held in the vehicle model performance specifications 21. The processor 1302 stores the calculated characteristic values ​​of each vehicle model (as vehicle model characteristic values ​​31) in the storage 1304. The process of calculating the characteristic values ​​of the vehicle models corresponds to the process by the vehicle model characteristic calculation unit 11 described above.

[0050] The characteristic values ​​of each vehicle model are data in which the performance specification data of each vehicle model has been reduced to a feature vector of a predetermined dimension (a vector expressed in a lower dimension than the performance specification data but with a dimension sufficient to retain the original information to some extent. "Retaining a certain degree" means that characteristic differences remain from the trends in the performance specification data of the vehicle models used as a population. "Characteristic differences" refers to a reduction in dimension sufficient to retain, for example, the perspectives that users prioritize when selecting a vehicle.) In other words, the performance specification data of each vehicle model is mapped by a machine learning model into a vector space of a predetermined dimension (as the characteristic values ​​of each vehicle model) that allows accurate comparison of the characteristics of individual vehicle models. Furthermore, by being mapped by the machine learning model, the vector space of the characteristic values ​​of each vehicle model can be a space that reflects (weights) the vehicle model's performance (e.g., horsepower or price) that is likely to be prioritized when purchasing a vehicle. In this embodiment, the vectorized characteristic values ​​of each vehicle model enable meaningful calculations between the characteristic values ​​of multiple vehicle models. For example, by calculating the average value (average vector) of the characteristic values ​​of multiple vehicle models, the average value represents the average characteristics of the multiple vehicle models. Furthermore, the difference between the characteristic value of a specific vehicle model and the average characteristic value of the multiple vehicle models represents the difference between the characteristic of the specific vehicle model and the average characteristic of the multiple vehicle models. Conventionally, when vehicle performance specification data is used directly as characteristic values, it tends to become a high-dimensional vector, and small differences in the performance specification data accumulate, making it difficult to compare vehicle models based on their characteristic differences. On the other hand, in this embodiment, by using data (vehicle characteristic values) in which the performance specification data of each vehicle model is mapped to a vector of a predetermined dimension, highly accurate recommendations based on data that emphasizes the characteristic differences between vehicle models are made possible.

[0051] The following process is performed for a specific user. Therefore, when recommending car models to multiple users, the process described below may be repeated for each user.

[0052] In S1401, the processor 1302 acquires the attributes of a specific user at a predetermined time, information on the vehicle model owned at the predetermined time, and current attributes. For example, the processor 1302 acquires this information of the specific user input via an information processing terminal included in the automobile recommendation system 1 via the communication interface 1303. Alternatively, the processor 1302 may acquire the user identifier input via the information processing terminal, and acquire this information of the specific user from the user's vehicle model 22, user attributes at the time of ownership start 23, and current user attributes 24.

[0053] In S1402, the processor 1302 determines a first group of users who belong to attributes that are the same as or similar to the attributes of the specific user at the predetermined time. For example, the processor 1302 determines users who have the same attributes (e.g., gender, residential area, and age group) as the specific user at the predetermined time as the users of the first group. The processor 1302 searches for the users of the first group, for example, from the user attributes at the start of ownership 23. The attributes of the specific user at the predetermined time may further include household composition and hobbies.

[0054] In S1403, the processor 1302 calculates characteristic values ​​of vehicle models requested by the first group of users. For example, the processor 1302 acquires the vehicle models requested by each user of the first group from the user-owned vehicle models 22, and then acquires the characteristic values ​​of the vehicle models requested by each user of the first group using the vehicle model characteristic values ​​31. Furthermore, the processor 1302, for example, calculates the average value of the acquired characteristic values ​​of each vehicle model, and sets this as the characteristic value of the vehicle model requested by the first group of users. The processing of S1402 and S1403 corresponds to the processing of the user attribute-requested vehicle model characteristic calculation unit 13. The processor 1302 may store the characteristic values ​​of the vehicle models requested by the first group of users in the user attribute-requested vehicle model characteristics 33. The processing of S1402 and S1403 makes it possible to treat the characteristics of vehicle models requested by multiple users with specific attributes as a single vectorized characteristic value.

[0055] In S1404, the processor 1302 acquires the characteristic values ​​of the vehicle model requested by the specific user at the specified time from the information on the vehicle models owned at the specified time. This processing corresponds to the processing of the user-requested vehicle model characteristic calculation unit 12. The processor 302 acquires the characteristic values ​​of the vehicle model requested by the specific user at the specified time from the information on the vehicle models owned by the specific user at the specified time and the per-vehicle characteristic values ​​31. The processor 1302 may store the characteristic values ​​of the vehicle model for the specific user in the per-user requested vehicle model characteristic 32.

[0056] In S1405, the processor 1302 calculates the difference between the characteristic values ​​of the vehicle models requested by the first group of users and the characteristic values ​​of the vehicle models requested by a specific user at a predetermined time, and stores the difference (in the per-user value value 34) as an index indicating the user's values. This processing corresponds to the processing of the per-user value value calculation unit 14. The calculated difference is a vectorized value of a predetermined dimension, and is therefore expressed as a calculable quantity indicating the directionality and magnitude of the user's values ​​in a space related to the vehicle model characteristics.

[0057] In S1406, the processor 1302 determines users in the second group who belong to attributes identical or similar to the current attributes of the specific user. For example, the processor 1302 determines users whose current attributes (e.g., gender, residential area, and age category) match those of the specific user as users in the second group. In S1407, the processor 1302 adds the characteristics of the vehicle model requested by the users in the second group to an index indicating the specific user's values ​​to calculate a combined vehicle model characteristic value (corresponding to the vehicle model characteristics currently requested by the specific user). For example, the processor 1302 obtains the vehicle model characteristics requested by the users in the second group (from the requested vehicle model characteristics per user attribute 33) and adds them to an index indicating the specific user's values. The value obtained by the addition (characteristic value) is a vectorized value of a predetermined dimension, so the vehicle model characteristics currently requested by the specific user are expressed in a space related to vehicle model characteristics. The processing of S1406 and S1407 corresponds to the processing of the current-user-requested vehicle model characteristic calculation unit 15. The processor 1302 may store the value (characteristic value) obtained by the addition in the current user requested vehicle model characteristic 35 .

[0058] In S1408, the processor 1302 determines a candidate vehicle model (e.g., one with similar characteristics) based on the combined characteristic value of the vehicle model and the characteristic value of each vehicle model (actual characteristic value present in the vehicle model characteristic values ​​31). For example, the processor 1302 compares the combined characteristic value of the vehicle model with the characteristics of each vehicle model, and extracts, for example, one or more characteristic values ​​of the vehicle model (characteristic values ​​of actual vehicle models similar to the combined characteristic value of the vehicle model) that fall within a predetermined Euclidean distance. The processor 1302 then determines the candidate vehicle model corresponding to the extracted characteristic value of the vehicle model as the vehicle model to be recommended (candidate vehicle model). Note that if there are multiple candidate vehicle models, the candidate vehicle models may be sorted so that the higher the similarity to the combined characteristic value of the vehicle model, the higher the priority, and the vehicle model with the higher priority may be displayed at the top.

[0059] In S1409, the processor 1302 recommends the candidate vehicle models to the specific user. The processes of S1408 and S1409 correspond to the processes of the recommended vehicle model calculation unit 16. For example, the processor 1302 may provide a list of candidate vehicle models to an information processing terminal and display the list on a display of the information processing terminal. After executing the process of S1409, the processor 1302 ends the series of operations of this process.

[0060] As described above, the automobile recommendation system 1 according to one embodiment of the present invention can recommend automobiles by inferring differences and changes in needs due to differences in user attributes and the values ​​of individual users based on the performance indexes of the vehicle models owned by the users. By separating the trends of each user's attributes from the values ​​of individual users in this way, automobiles can be recommended according to changes in life stages.

[0061] The invention made by the inventor has been specifically described above based on the embodiments, but it goes without saying that the present invention is not limited to the above embodiments and can be modified in various ways without departing from the spirit of the invention. Furthermore, the above embodiments have been described in detail to clearly explain the present invention, and the present invention is not necessarily limited to those having all of the described configurations. Furthermore, it is possible to add, delete, or replace part of the configuration of the above embodiments with other configurations.

[0062] Furthermore, the above-described configurations, functions, processing units, processing means, etc. may be partially or entirely implemented in hardware, for example, by designing them as integrated circuits. The above-described configurations, functions, etc. may also be implemented in software, with a processor interpreting and executing a program that implements each function. Information such as the programs, tables, and files that implement each function can be stored in a storage device such as a memory, hard disk, or SSD, or on a storage medium such as an IC card, SD card, or DVD.

[0063] In addition, in the above figures, the control lines and information lines shown are those that are considered necessary for explanation, and do not necessarily show all the control lines and information lines that are actually implemented. In reality, it can be assumed that almost all components are interconnected.

[0064] The present invention can be used in a car recommendation system that recommends cars.

[0065] 1...Automobile recommendation system, 11...Vehicle model characteristic calculation unit, 12...User-requested vehicle model characteristic calculation unit, 13...Vehicle model characteristic calculation unit required for each user attribute, 14...User-requested value calculation unit, 15...Current user-requested vehicle model characteristic calculation unit, 16...Recommended vehicle model calculation unit, 21...Performance specifications for each vehicle model, 22...Vehicle model owned by user, 23...User attributes at start of ownership, 24...Current user attributes, 31...Characteristic value for each vehicle model, 32...Vehicle model characteristic required for each user, 33...Vehicle model characteristic required for each user attribute, 34...User-requested value, 35...Current user-requested vehicle model characteristic, 36...Recommended vehicle model

Claims

1. A car recommendation system that recommends cars to a user, comprising: a memory that stores instructions; and one or more processors, which, when the instructions are executed, cause the car recommendation system to do the following: determine a first group of users who belong to attributes that are the same as or similar to the attributes of a specific user at a given time; calculate an index indicating the user's values ​​from the difference between the characteristic values ​​of the car models requested by the users of the first group and the characteristic values ​​of the car models requested by the specific user at the given time; determine a second group of users who belong to attributes that are the same as or similar to the specific user's current attributes; calculate an aggregate characteristic value of the car models based on the characteristic values ​​of the car models requested by the users of the second group and the index indicating the specific user's values; determine candidate car models based on the aggregate characteristic value of the car models and the characteristic values ​​of each car model; and recommend the candidate car models to the specific user.

2. An automobile recommendation system as described in claim 1, wherein determining the candidate vehicle model includes determining the characteristic value of a vehicle model that is within a predetermined distance from the combined characteristic value of the vehicle model as the characteristic value of the candidate vehicle model, and determining a vehicle model that has the characteristic value of the candidate vehicle model as the candidate vehicle model.

3. An automobile recommendation system as described in claim 1, wherein determining the candidate vehicle models includes determining a plurality of candidate vehicle models to be displayed on a display based on the characteristics of the combined vehicle models and the characteristics of each vehicle model, and the plurality of candidate vehicle models are sorted so that the higher the similarity with the characteristic values ​​of the combined vehicle models, the higher the priority, and vehicle models with higher priority are displayed at the top.

4. An automobile recommendation system as described in claim 1, wherein the characteristic values ​​of the vehicle models are data in which the performance specifications of the vehicle models are mapped into a vector space in which the characteristics of individual vehicle models can be compared.

5. An automobile recommendation system as described in claim 1, wherein, when the command is executed, the automobile recommendation system uses a machine learning model to calculate characteristic values ​​of a vehicle model by mapping data on performance specifications of the vehicle model to a feature vector expressed in lower dimensions than the data on the performance specifications.

Images