Information processing device and information processing method

Abstract

An information processing device according to an aspect of the present disclosure comprises: an acquisition unit that acquires the feature amounts of a plurality of items of a plurality of stores belonging to a store group that is the object of processing among one or more store groups composed of a plurality of stores; and an output unit that outputs, for a singular item the value of store group feature amount of which is unique among the plurality of items of the store group feature amounts calculated from the feature amounts of the plurality of stores belonging to the store group that is the object of processing, data of feature amounts in which singularity is removed from the feature amounts of the singular item of the plurality of stores.

Description

Information processing device and information processing method

[0001] The present invention relates to a technology that takes into account information about the specificity of features of stores used for learning in a service that recommends stores to users.

[0002] Techniques for reducing data bias in data sets used in machine learning and the like are known. For example, Patent Document 1 discloses a system for clustering word features to enable highly accurate extraction of words with specific meanings from text. In this system, it is determined whether there is a bias in the feature vectors of each cluster, and if a bias is determined, each feature vector is updated to reduce the bias. Patent Document 2 discloses a system for changing the probability distribution of each component entity in a specific environment to reduce the bias in the training data, in order to solve the problem of inability to accurately reproduce the probability distribution of a training target due to bias in the training data.

[0003] Patent No. 5751251 Patent No. 7437738

[0004] Neither of the techniques described in Patent Documents 1 and 2 is intended for a service that recommends stores to users.

[0005] In response to this, the present invention provides a technology for providing information on store feature amounts of a store group having an item with a unique value, out of store group feature amounts consisting of multiple items in a store recommendation service.

[0006] One aspect of the present disclosure provides an information processing device that has an acquisition unit that acquires feature values ​​of multiple items of multiple stores belonging to a target store group among one or more store groups consisting of multiple stores, and an output unit that outputs information regarding the specificity of a unique item, among the multiple items of store group feature values ​​calculated from the feature values ​​of the multiple stores belonging to the target store group, whose value of the store group feature value is unique.

[0007] Another aspect of the present disclosure provides an information processing method including the steps of: acquiring feature values ​​for multiple items of multiple stores belonging to a target store group among one or more store groups consisting of multiple stores; and outputting information regarding the uniqueness of a unique item, for which the value of the store group feature is unique, from the feature values ​​of the multiple stores in the unique item calculated from the feature values ​​of the multiple stores belonging to the target store group.

[0008] According to the present invention, it is possible to provide information on store feature amounts of a store group having an item with a unique value, for a store group feature amount consisting of a plurality of items in a store recommendation service.

[0009] FIG. 1 is a diagram showing an overview of an information processing system 1 according to an embodiment. FIG. 2 is a diagram explaining a problem in a recommendation service. FIG. 3 is a diagram showing the concept of removal processing. FIG. 4 is a diagram illustrating the functional configuration of the information processing system 1. FIG. 5 is a diagram illustrating the hardware configuration of an information processing device 10. A flowchart showing the operation of the information processing system 1. FIG. 6 is a diagram illustrating a user feature amount CU. FIG. 7 is a diagram illustrating a store feature amount CP. FIG. 8 is a diagram illustrating a score indicating uniqueness. FIG. 9 is a diagram illustrating a method for generating a store feature amount CP. FIG. 10 is a diagram illustrating a graph G that is generated. FIG. 11 is a diagram illustrating recommendation processing in the information processing system 1.

[0010] 1. Overview FIG. 1 is a diagram illustrating an overview of an information processing system 1 according to an embodiment. The information processing system 1 provides a service (hereinafter referred to as a "recommendation service") that recommends a recommended store P from among multiple stores P to a user U. The information processing system 1 includes an information processing device 10 and multiple user terminals 20. The information processing device 10 is a server in the recommendation service. The user terminal 20 is a client in the recommendation service. The information processing device 10 and the user terminals 20 communicate via a network 9. The network 9 is a computer network and includes at least one of a mobile network and the Internet. The information processing system 1 provides a recommendation service to multiple users U. When distinguishing between the multiple users U, they are referred to as "missing user U[1]," "missing user U[2]," ..., "missing user U[k]," etc. Similarly, when distinguishing between the multiple stores P, they are referred to as "store P[1]," "store P[2]," ..., "store P[m]," etc.

[0011] It is preferable that the recommendation service recommends a store P that the user U would actually like to visit. One recommendation method is a method using machine learning. One example of a machine learning model used in this machine learning is a model that uses the features of the user U and the features of the store P as explanatory variables and the visit history as a target variable. In this way, three types of information are required to develop a machine learning model: the features of the user U, the features of the store P, and the visit history.

[0012] Here, the feature amount of user U is referred to as user feature amount CU, and the feature amount of store P is referred to as store feature amount CP (user feature amount CU and store feature amount CP are not shown in Figure 1). Both user feature amount CU and store feature amount CP are vectors. The features of user U or store P are divided into multiple items. Each component of these vectors quantifies the value of each item. Although these vectors are the information that forms the basis for store P recommendations, the feature amounts of similar stores tend to be similar. As a result, a situation may arise in which user U does not obtain the recommendation results he or she desires.

[0013] Figure 2 illustrates the challenges faced by recommendation services. The store feature C is an n-dimensional vector, but for simplicity, we will consider an example where n=2. Figure 2 plots the store feature C on a two-dimensional plane (xy plane). Because the store feature C of stores P with a common attribute tends to take similar values, the plot of the store feature C for multiple stores P forms clusters on the xy plane based on the attribute. In the example of Figure 2, three clusters (cluster A, cluster B, and cluster C) are formed. These clusters are formed, for example, by industry. In this example, cluster A is a cluster of hot spring facilities, cluster B is a cluster of restaurants, and cluster C is a cluster of souvenir shops. Of these, cluster A includes stores P[1] and P[2].

[0014] Consider an example in which a hot spring guide app uses this recommendation service in such a situation. The hot spring guide app is an app that provides information on multiple hot spring facilities. By using the recommendation service, the hot spring guide app is expected to recommend a facility that matches the user's preferences from among multiple hot spring facilities. A recommendation service targeting hot spring facilities requires selecting a facility that matches the user's preferences from a group of stores P with similar features, called cluster A. However, because the features of the stores P belonging to cluster A are similar, differentiation is difficult, and there are cases in which recommendations cannot be made successfully. In this embodiment, a technology is provided that makes it easier to differentiate between stores P within a cluster by performing a process (hereinafter referred to as "removal process") to remove factors that cause multiple stores P to form a cluster, i.e., factors that hinder variation.

[0015] FIG. 3 is a diagram illustrating the concept of the elimination process. For simplicity, we will focus on a specific feature ("age" in the example shown). In FIG. 3, the horizontal axis represents age, and the vertical axis represents the number of stores P. The graph on the left of FIG. 3 shows the distribution of the features of store P in cluster A before the elimination process. In this example, age is not normally distributed, but has a biased distribution with a sharp peak at the relatively high age of Yp years. The graph on the right of FIG. 3 shows the distribution of the features of store P in cluster A after the elimination process. The elimination process has smoothed out the bias in the distribution, making it closer to a normal distribution. By bringing the distribution closer to a normal distribution in this way, it becomes easier to make recommendations using the value of that item.

[0016] 4 is a diagram illustrating an example of the functional configuration of the information processing system 1. The information processing system 1 includes an acquisition unit 11, a processing unit 12, an output unit 13, a learning unit 17, a storage unit 18, and a control unit 19. In this example, the acquisition unit 11, the processing unit 12, the output unit 13, the learning unit 17, the storage unit 18, and the control unit 19 are implemented in an information processing device 10.

[0017] The acquisition unit 11 acquires store features CP of multiple stores P belonging to a target store group GP. The store group GP is a collection of multiple stores P and corresponds to the cluster described in the example of Figure 2. The target store group is a group of stores designated as targets among the store group GP. As described above, the store features CP have multiple items. The processing unit 12 performs a process of identifying peculiar items and a process of removing the peculiar items. A peculiar item is an item, among the multiple items of the store features CP, whose value in the store group GP is peculiar. Here, "a value that is peculiar in the store group GP" means that the distribution of values ​​deviates from a normal distribution. The removal process is a process of removing the peculiarities of the peculiar items from the store features CP of multiple stores P belonging to the store group GP. The output unit 13 outputs information regarding the peculiarities of the peculiar items of the features of the multiple stores P belonging to the target store group GP. In this example, the "information regarding specificity" includes data on features from which specificity has been removed for specific items in the features of multiple stores P belonging to the target store group GP ("store features (revised)" in the figure).

[0018] The learning unit 17 trains the machine learning model M using the data of the store feature CP from which the idiosyncrasies have been removed and output from the output unit 13. When input data corresponding to a user (e.g., user U feature) is provided to the trained machine learning model M, it outputs information about a store to be recommended to that user (e.g., store ID). The information processing device 10 can provide a recommendation service using the machine learning model M trained in this way. The memory unit 18 stores various data and programs. The control unit 19 performs various controls.

[0019] FIG. 5 is a diagram illustrating an example of the hardware configuration of the information processing device 10. Physically, the information processing device 10 is configured as a computer including a processor 101, a memory 102, a storage 103, a communication device 104, and a bus connecting these devices. Each of these devices operates using power supplied from a battery (not shown). In the following description, the term "device" can be interpreted as a circuit, device, unit, or the like. The hardware configuration of the information processing device 10 may be configured to include one or more of the devices shown in FIG. 3, or may be configured without including some of the devices. Furthermore, the information processing device 10 may be configured by communicating with multiple devices each having a different housing.

[0020] Each function of the information processing device 10 is realized by loading specified software (programs) onto hardware such as the processor 101, memory 102, etc., so that the processor 101 performs calculations, controls communication via the communication device 104, and controls at least one of reading and writing data in the memory 102 and storage 103.

[0021] The processor 101 controls the entire computer by running, for example, an operating system. The processor 101 may be configured as a central processing unit (CPU) including an interface with peripheral devices, a control device, an arithmetic unit, a register, etc. Furthermore, for example, a baseband signal processing unit, a call processing unit, etc. may be realized by the processor 101.

[0022] The processor 101 reads programs (program codes), software modules, data, etc. from at least one of the storage 103 and the communication device 104 into the memory 102 and executes various processes in accordance with these. The programs used are those that cause a computer to execute at least some of the operations described below. The functional blocks of the information processing device 10 may be implemented by a control program stored in the memory 102 and running on the processor 101. Various processes may be executed by one processor 101, or may be executed simultaneously or sequentially by two or more processors 101. The processor 101 may be implemented by one or more chips. The programs may be transmitted to the information processing device 10 via a telecommunications line.

[0023] The memory 102 is a computer-readable recording medium and may be configured by, for example, at least one of a read-only memory (ROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), a random access memory (RAM), etc. The memory 102 may also be called a register, a cache, a main memory (primary storage device), etc. The memory 102 can store executable programs (program codes), software modules, etc. for implementing the method according to this embodiment.

[0024] Storage 103 is a computer-readable recording medium, and may be composed of at least one of, for example, an optical disk such as a CD-ROM (Compact Disc ROM), a hard disk drive, a flexible disk, a magneto-optical disk (e.g., a compact disk, a digital versatile disk, a Blu-ray (registered trademark) disk), a smart card, a flash memory (e.g., a card, a stick, a key drive), a floppy (registered trademark) disk, a magnetic strip, etc. Storage 103 may also be called an auxiliary storage device.

[0025] The communication device 104 is hardware (transmission / reception device) for communicating between computers via at least one of a wired network and a wireless network, and is also called, for example, a network device, a network controller, a network card, or a communication module.

[0026] Each device, such as the processor 101 and the memory 102, is connected by a bus for communicating information. The bus may be configured using a single bus, or different buses may be used between each device.

[0027] The information processing device 10 may be configured to include hardware such as a microprocessor, a digital signal processor (DSP), an application specific integrated circuit (ASIC), a programmable logic device (PLD), or a field programmable gate array (FPGA), and some or all of the functional blocks may be realized by the hardware. For example, the processor 101 may be implemented using at least one of these pieces of hardware.

[0028] In this example, the programs stored in the storage 103 include a program (hereinafter referred to as a "server program") for causing the computer to function as the information processing device 10 (i.e., a server) in the information processing system 1. When the processor 101 is executing the server program, the processor 101 is an example of the acquisition unit 11, the processing unit 12, the learning unit 17, and the control unit 19. The communication device 104 is an example of the output unit 13.

[0029] 6 is a flowchart showing the operation of the information processing system 1. Before the start of the processing in FIG. 6, the information processing system 1 or another system generates user feature quantities CU for multiple users U and store feature quantities CP for multiple stores P. Any specific method may be used to generate the user feature quantities CU and the store feature quantities CP.

[0030] FIG. 7 is a diagram illustrating an example of a user feature CU. In one example, the user feature CU is (automatically) generated from subscriber information (or user features) registered with a telecommunications carrier. The subscriber information is information indicating the characteristics or attributes of a user, and includes values ​​for multiple items, such as the user U's gender, age group, place of residence, hobbies, and marital status (top row of FIG. 7). The user feature CU is a normalized value of each item of the subscriber information. Normalization refers to converting the value to a predetermined numerical value, for example, a real number between 0 and 1 (according to a predetermined rule) (bottom row of FIG. 7). This conversion may be performed by the information processing device 10, or the information processing device 10 may acquire data converted by an external system. Because the subscriber information includes multiple items, the user feature CU is a vector whose components are the values ​​of each item.

[0031] The illustrated user feature value CU is the user feature value CU of each of the multiple users U. For example, the example in the top row of Fig. 7 indicates that the gender of user U[1] is "1" (indicating female), the age is "0.2" (indicating that the user is in his / her twenties), the place of residence is "0.13" (indicating that the user lives in Tokyo), and the hobby is "0.40" (indicating that the user is traveling).

[0032] FIG. 8 is a diagram illustrating an example of a store feature CP. In one example, the store feature CP is manually input by, for example, a store operator or an investigator. In another example, the store feature CP is automatically generated using machine learning from visit history information indicating a user U's visits to the store P and the user feature CU. The store feature CP includes values ​​for each of multiple items. These values ​​are normalized. In this example, the store feature CP has the same items as the user feature CU, namely, gender, age group, place of residence, hobbies, and marital status. This can be interpreted as representing the characteristics of customers who frequently visit the store P. The store feature CP may be generated by the information processing device 10, or the information processing device 10 may acquire a store feature CP generated by another system.

[0033] The multiple stores P are classified into multiple store groups GP, i.e., clusters, based on the store feature values ​​CP. Any known clustering method can be used, such as the k-means method, the shortest distance method, the furthest distance method, the group average method, or Ward's method. The store feature values ​​CP are clustered, and each store feature value CP is associated with a cluster ID that specifies which store group GP the store P belongs to. The clustering may be performed by the information processing device 10, or the information processing device 10 may acquire the results of clustering performed by an external system.

[0034] Referring again to FIG. 6 , in step S1, the information processing device 10 acquires store features CP for stores P belonging to a certain population (hereinafter referred to as the “target population”). Prior to acquiring the store features CP, the information processing device 10 identifies the target population. The target population includes one or more store groups GP. In one example, the target population is predetermined by a server program. Specifically, the target population is, for example, a set of all stores P for which the information processing device 10 can acquire store features CP. In another example, the target population is specified by an application using the recommendation service. In this case, the target population is a set of stores P narrowed down by some filtering condition from among the stores P for which the information processing device 10 can acquire store features CP. The filtering condition is a condition related to the attributes or store features CP of the store P, for example, a condition related to at least one of the location, business type, and average customer spending. Alternatively, the narrowing down may be performed based on the store group GP, i.e., the cluster ID. When narrowing down the stores P, clustering is performed on the target population, i.e., the narrowed down population. Alternatively, clustering may be performed on the original population (i.e., the large population), and the target population may be narrowed down from the large population. Specifically, if the application using the recommendation service is a hot spring guide app, the hot spring guide app issues an instruction to narrow down the target population to hot spring facilities. The information processing device 10 narrows down the target population to hot spring facilities in accordance with the instruction from the hot spring guide app.

[0035] In step S2, the information processing device 10 identifies the store group GP included in the target population. As illustrated in FIG. 8 , the store feature CP is associated with a cluster ID. The information processing device 10 can identify the store group GP included in the target population using the cluster ID. Specifically, the information processing device 10 identifies the number of store groups GP included in the target population. If there is only one type of cluster ID corresponding to the acquired store feature CP, the target population will contain one store group GP. If there are multiple types of cluster IDs corresponding to the acquired store feature CP, the target population will contain multiple store groups GP.

[0036] In step S3, the information processing device 10 identifies peculiar items for each of the identified store groups GP. To identify the peculiar items, a score indicating the peculiarities of the values ​​of each item of the store group feature amount CG is used. The store group feature amount CG is the store feature amount of the store group obtained from the store feature amount CP of a store P belonging to the target store group GP.

[0037] FIG. 9 is a diagram illustrating scores indicating uniqueness. In this example, scores are shown for multiple items of the store feature CP, namely, gender, age group, place of residence, hobbies, and marital status. In this example, the score is the variance of the values ​​for each item. The smaller the variance, the closer the distribution is to the mean value. This indicates that the ages of customers at that store P are biased toward a specific value, making it unique. Conditions for determining whether an item is unique are defined for this score. A value that is likely to satisfy the condition for becoming a unique item is called a "high" score, and a value that is difficult to satisfy is called a "low" score. In the example of variance, the smaller the numerical value, the higher the score, and the larger the numerical value, the lower the score. In one example, this condition is that an item whose variance is below a threshold is considered unique. This threshold is predefined. The information processing device 10 determines whether the variance satisfies this condition. The information processing device 10 identifies an item whose score satisfies the condition as a unique item. If none of the item scores satisfy the condition, the information processing device 10 determines that there are no peculiar items for the store group GP. In the example of Figure 9, among the items of the store group feature CG[1], the variance of "age group" is below a defined threshold (e.g., 0.15). In this example, age group is identified as an peculiar item for the store group GP[1].

[0038] Referring again to FIG. 6 , in step S4, the information processing device 10 performs a removal process. The removal process is performed on data of items corresponding to peculiar items among the store group features CG determined to contain peculiar items. In the example of FIG. 9 , the removal process is performed on data of the item "age group" among the store features CP of the store group GP[1]. The removal process is not performed on data of other items. The removal process is a transformation process for approximating the distribution of data of items corresponding to peculiar items to a normal distribution, and includes a Box-Cox transformation in one example. The Box-Cox transformation is a transformation expressed by the following equation (1): Here, λ is a coefficient, and a value (for example, an integer) is selected using, for example, the Shapiro-Wilk test, such that the distribution of the transformed data is closest to a normal distribution.

[0039] The information processing device 10 may repeatedly perform the removal process on the data of the target store group GP until the defined convergence condition is satisfied.

[0040] In step S5, the information processing device 10 stores the store feature CP including the data after the removal process. This process can also be described as the information processing device 10 updating or overwriting the original store feature CP data with the data after the removal process. Thereafter, the information processing system 1 can provide a recommendation service using the updated store feature CP. More specifically, the information processing system 1 can use the updated store feature CP as training data to train a machine learning model, or can use the updated store feature CP to match with the user feature CU.

[0041] 3-2. Score indicating specificity The score indicating specificity is not limited to the above example. This score may be, for example, item importance. Item importance is an index indicating the magnitude of contribution to the store feature amount CP among multiple items that make up the store feature amount CP. An example of a method for calculating item importance will be described below, but as a preparation for that description, a method for generating the store feature amount CP will first be described.

[0042] FIG. 10 is a diagram illustrating a method for generating a store feature C. In step S21, the information processing device 10 acquires visit history information. In step S22, the information processing device 10 acquires a user feature C. In this example, a system for generating visit history information and a system for generating user feature C exist separately from the information processing system 1, and the information processing system 1 acquires the user feature C and visit history information from these systems. The visit history information indicates a visit history to a specified target store during a specified period. In step S23, the information processing device 10 generates a graph of a graph neural network GNN based on the visit history information.

[0043] FIG. 11 is a diagram illustrating an example of a generated graph G. Graph G has multiple nodes and multiple links (also called edges). The nodes are classified into user nodes and store nodes. A user node is a node corresponding to a user U. A store node is a node corresponding to a store P. Links represent visit history. A link connecting user node [1] and store node [1] indicates that user U [1] has a history of visiting store P [1]. There is no link between a user U with no visit history and a store P. In the following description, a user U who has a link with a store P from the perspective of a certain store P is referred to as a "neighboring user." Similarly, a store P that has a link with the user U from the perspective of a certain user U is referred to as a "neighboring store."

[0044] FIG. 11 illustrates a graph consisting of four users U[1] to U[4] and three stores P[1] to P[3].

[0045] Returning to FIG. 10 , in step S24, the information processing device 10 initializes the store feature CP of each store P. In this example, the store feature CP is a vector with the same number of components as the user feature CU. The initial value of the store feature CP is predetermined, for example, all items are null. In step S25, the information processing device 10 updates the store feature CP. When updating the store feature CP of a certain store P, a calculation is performed to convolve the user feature CU of users surrounding the store P. Convolving the user feature CU means weighting the user feature CU of the surrounding users and applying an aggregation function to these (for example, calculating the average value). The information processing device 10 updates the store feature CP for all store nodes included in the graph G. In step S26, the information processing device 10 updates the user feature CU. When updating the user feature CU of a certain user U, a calculation is performed to convolve the store feature CP of stores surrounding the user U. Convolving the store feature C means weighting the store feature C of the surrounding stores and applying an aggregation function to them (for example, calculating the average value). The information processing device 10 updates the user feature C for all user nodes included in the graph G.

[0046] Returning to FIG. 10 , in step S27, the information processing device 10 determines whether a convergence condition is satisfied for the store feature C. The convergence condition is predetermined. The convergence condition may be, for example, a condition related to the match rate between the value obtained by applying a predetermined activation function to the distance between each user feature C and the store feature C (i.e., the distance between the two vectors) and the visit history information for the combination of the user U and the store P. Specifically, the convergence condition is that the amount or rate of change in this match rate is equal to or less than a threshold. As the activation function, for example, a well-known function such as a sigmoid function or a hyperbolic tangent function (tanh) is used.

[0047] If the convergence condition is satisfied (step S27: YES), the information processing device 10 proceeds to step S28. If the convergence condition is not satisfied, the information processing device 10 proceeds to step S25 again. In this way, the processes of steps S25 to S26 are repeatedly executed until the convergence condition is satisfied.

[0048] In step S28, the information processing device 10 outputs the converged store feature quantity C. In one example, the information processing device 10 writes the value of the converged store feature quantity C into a store feature quantity C database. In this way, the information processing device 10 can obtain the store feature quantity C without information input by the manager or investigator of the store P.

[0049] Assuming that the store feature CP is generated as described above, the information processing device 10 temporarily changes the value of one target item (herein referred to as the "target item") among the multiple items included in the user feature CU to a specific value (e.g., zero). The information processing device 10 calculates the store feature CP using the user feature CU whose value has been changed in the same manner as described above. The store feature CP calculated here is referred to as the "temporary store feature CP." The information processing device 10 calculates the vector distance between the original store feature CP and the temporary store feature CP. The information processing device 10 calculates the temporary store feature CP by sequentially determining the multiple items as target items one by one, and calculates the vector distance between the temporary store feature CP and the original store feature CP. The information processing device 10 updates the target item to the next item and similarly calculates the vector distance between the updated target item and the original store feature CP. The information processing device 10 repeats this process until all items have been treated as target items. The information processing device 10 normalizes the inter-vector distance calculated for each item in this way and stores it as the item importance of that item.

[0050] As can be seen from this calculation method, the larger the item importance value, the greater the contribution to the store feature CP. In other words, the larger the item importance value, the higher the score, and the smaller the item importance value, the lower the score. The information processing device 10 determines that an item whose calculated item importance value exceeds a threshold value is an anomalous item.

[0051] 3-3. Removal Process The removal process is not limited to the above example. The removal process may be a conversion process using a weight vector. This process is performed as follows. First, the information processing device 10 takes all users U as a population and calculates the average value and standard deviation of each item of the user feature amount CU of that population. Next, the information processing device 10 takes users U corresponding to the target store group GP as a population and calculates the average value and standard deviation of each item of the users U of that population. The information processing device 10 calculates a weight vector Vw using the calculated average value. Any specific definition of the weight vector may be used. For example, In this case, when Ave(Usa[i]) is greater than Ave(Uall[i]), a weight is assigned that reduces Ave(Usa[i]) after calculation using equation (3) described below, and when Ave(Usa[i]) is smaller than Ave(Uall[i]), a weight is assigned that increases Ave(Usa[i]) after calculation. Here, Ave(Usa[i]) indicates the average value of the i-th item of user features, with the user U corresponding to the target store group GP as the population. The user U corresponding to the target store group GP refers to, for example, a group of users U who have a history of visiting the target store group GP. Ave(Uall[i]) indicates the average value of the i-th item of user features, with all users U as the population. Th indicates a defined threshold.

[0052] Next, the information processing device 10 applies the obtained weight vector Vw to the user feature CU of the user U corresponding to the target store group GP, i.e., multiplies it by each component (i.e., each item) (Equation (3)). The information processing device 10 obtains the corrected user feature CUr through this calculation.

[0053] Next, the information processing device 10 recalculates the store feature amount CP from the obtained user feature amount CUr. The method for calculating the store feature amount CP from the user feature amount CU is as described in Fig. 10. In this way, the information processing device 10 obtains the store feature amount CP from which the idiosyncrasies of the peculiar items have been removed.

[0054] The information processing device 10 may be capable of executing multiple processes, including a process using a distribution transformation such as a Box-Cox transformation and a process using a weight vector. In this case, the information processing device 10 may determine which process to use as the removal process depending on the distribution of the store feature amount C in the target store group GP. In one example, the information processing device 10 uses a process using a distribution transformation as the removal process when the distribution of the store feature amount C is biased. Furthermore, the information processing device 10 uses a process using a weight vector as the removal process when the average value of the store feature amount C is higher or lower than a reference value.

[0055] To determine whether or not there is bias in the distribution of the store feature Cp, for example, the Shapiro-Wilk test is used. The Shapiro-Wilk test tests the null hypothesis that the samples x1, ..., xn have been sampled from a population that follows a normal distribution. In the Shapiro-Wilk test, if the statistic W obtained by the following formula is smaller than a criterion (for example, 0.05 or 0.01), the null hypothesis is rejected.

[0056] 12 is a diagram illustrating a recommendation process in the information processing system 1. In step S11, the user terminal 20 transmits a recommendation request to the information processing device 10. A client application program for a recommendation service is installed in the user terminal 20, and this application transmits the recommendation request. The recommendation request includes information identifying a user feature CU of a user U of the user terminal 20. The information identifying the user feature CU is, for example, the user feature CU itself or a user ID.

[0057] Upon receiving a recommendation request from the user terminal 20, the information processing device 10 inputs the user feature CU and the store feature CP into the machine learning model M (step S12). The machine learning model M is stored in the storage unit 18 of the information processing device 10; however, for simplicity, the machine learning model M is depicted as a separate element from the information processing device 10. The machine learning model M is a trained model that has undergone machine learning using training data in which the user feature CU and the store feature CP are used as explanatory variables (or inputs) and the visit history information is used as the objective variable (or output). The information processing device 10 identifies the user feature CU and the store feature CP from the recommendation request. For example, the recommendation request includes a user ID and conditions for narrowing down the stores (e.g., region or industry). The information processing device 10 has a user database in which user feature values ​​are recorded and a store database in which store feature values ​​are recorded (both not shown). The information processing device 10 refers to the user database to identify the user feature CU corresponding to the user ID included in the recommendation request. Furthermore, the information processing device 10 refers to the store database to identify the store feature amount CP of the store P narrowed down by the recommendation request.

[0058] When the user feature values ​​CU and store feature values ​​CP are input, the machine learning model M infers (or selects) the store P that is optimal for the user U from the input user feature values ​​CU and store feature values ​​CP (step S13). The machine learning model M outputs the store ID (i.e., identification information) of the store P (step S14). The information processing device 10 extracts attribute information of the store corresponding to the store ID output from the machine learning model M from the store database. The information processing device 10 transmits a recommendation result to the user terminal 20 (step S15). This recommendation result includes the identification information and other attribute information of the selected store P.

[0059] When receiving the recommendation result from the information processing device 10, the user terminal 20 displays information about the recommended store P (step S16). In this way, the user U can receive a recommendation of a store P that matches his or her own characteristics.

[0060] 4. Modifications The present invention is not limited to the above-described embodiment, and various modifications are possible. Some modifications will be described below. Two or more of the following features may be applied in combination.

[0061] (1) Anomalous Items The conditions for identifying anomalous items are not limited to those exemplified in the embodiment. In one example, a condition may be used such that an item with a large amount of change compared to other store groups is identified as an anomalous item. First, the information processing device 10 calculates the average value of each item of the store feature amount for the entire population. Next, the information processing device 10 calculates the value of each item of the store feature amount for the target store group (for example, the average value of the store feature amount for stores belonging to the target store group). The information processing device 10 subtracts the value of each item of the store feature amount for the entire population from the value of each item of the store feature amount for the target store group. The information processing device 10 identifies items for which the magnitude of the subtraction result exceeds a threshold as anomalous items.

[0062] Regardless of the conditions used to identify anomalous items, there may be multiple anomalous items in a single store group. For example, when a condition is used that identifies an item with a large amount of change compared to other store groups as an anomalous item, if there are multiple items whose subtraction results exceed a threshold, the information processing device 10 identifies all of the items whose values ​​exceed the threshold as anomalous items. In this case, the information processing device 10 may process items by distinguishing them as primary anomalous items, secondary anomalous items, and so on, in order of the strength with which the condition is satisfied (for example, the magnitude of the subtraction result).

[0063] The information processing device 10 may use different conditions for identifying peculiar items depending on the number of target store groups. For example, if there are multiple target store groups, the information processing device 10 uses the condition that "if the score of an item in a store group is higher than the score of that item in other store groups, then that item in that store group is designated as a peculiar item." "Higher than the score in other store groups" means that the difference between the score of an item in a store group and the average score of that item in the other store groups exceeds a threshold. If there is only one target store group, then the condition that "the item with the highest score among multiple items in the store group feature is designated as a peculiar item" is used. Note that the term "score" here encompasses the variance (the smaller the numerical value, the higher the score) and importance (the larger the numerical value, the higher the score) described in the embodiment.

[0064] (2) Removal Process The removal process is not limited to the example described in the embodiment. When a distribution transformation is used, a transformation process other than the Box-Cox transformation, such as a Yeo-Johnson transformation, a logarithmic transformation, a square root transformation, an arcsine transformation, or an exponential transformation, may be used.

[0065] (3) Store In the information processing system 1, the term "store" can be interpreted in a broad sense. For example, a store is not limited to a physical store (real store) and may be an online store (or virtual store) that does not have a physical entity. An online store may also be something that does not include the word "store," such as a website or an application program. For example, a "store" targeted by the information processing system 1 may include a website for downloading a specific application program.

[0066] (4) Store Feature Amount The store feature amount is not limited to the examples given in the embodiment. The store feature amount may not have the same items as the user feature amount, and may have items that are defined independently of the user feature amount.

[0067] (5) Output Data and Use The data output by the output unit 13 and its use are not limited to those exemplified in the embodiment. For example, instead of or in addition to outputting updated store feature data, the output unit 13 may output information identifying which of multiple items is an anomalous item (another example of information regarding the anomalous item for which the value of the store group feature is anomalous). The information processing system 1 may output this information for the purpose of utilizing it in store design or service design. For example, if a result is obtained that loss aversion tendency is an anomalous item for a store group targeting a certain store or product (e.g., an application program), the information processing device 10 outputs information to that effect. A user who receives this information can take measures in accordance with the anomalousness of the store feature, such as adding loss aversion language to recommendation text or advertising text. In another example, if a result is obtained that savings tendency is an anomalous item for a store group targeting a certain store or product (e.g., an application program), the information processing device 10 outputs information to that effect. A user who receives this information (e.g., a store operator) can implement measures such as discount campaigns for customers.

[0068] (6) System Configuration The system configuration of the information processing system 1 is not limited to that exemplified in the embodiment. At least some of the functions described in the embodiment as being implemented in the information processing device 10 may be implemented in another device such as the user terminal 20. Alternatively, at least some of the functions described in the embodiment as being implemented in another system may be implemented in the information processing device 10.

[0069] Furthermore, the correspondence between the functional configuration and the hardware configuration is not limited to that exemplified in the embodiment, and the information processing system 1 may be configured with devices having any hardware configuration as long as the required functions can be realized.

[0070] (7) Others The various programs executed by the processor 101 may be provided by downloading via a network such as the Internet, or may be provided in a state recorded on a computer-readable non-transitory recording medium such as a DVD-ROM. Each processor may be, for example, a CPU, an MPU (Micro Processing Unit), or a GPU (Graphics Processing Unit).

[0071] The block diagrams used to explain the above embodiments show functional blocks. These functional blocks (components) are realized by any combination of hardware and / or software. Furthermore, the method for realizing each functional block is not particularly limited. That is, each functional block may be realized using a single device that is physically or logically coupled, or may be realized using two or more physically or logically separated devices that are directly or indirectly connected (e.g., wired, wireless, etc.) and these multiple devices. The functional block may also be realized by combining software with the single device or multiple devices.

[0072] Functions include, but are not limited to, judgment, determination, judgment, calculation, computation, processing, derivation, investigation, search, confirmation, reception, transmission, output, access, resolution, selection, selection, establishment, comparison, assumption, expectation, consideration, broadcasting, notifying, communicating, forwarding, configuring, reconfiguring, allocating, mapping, and assignment. For example, a functional block (component) that performs transmission is called a transmitting unit or transmitter. As mentioned above, there are no particular limitations on how these functions are implemented.

[0073] For example, the information processing device 10 according to an embodiment of the present disclosure may function as a computer that performs the processing of the present disclosure.

[0074] Each aspect / embodiment described in the present disclosure may be applied to at least one of systems using LTE (Long Term Evolution), LTE-Advanced (LTE-A), SUPER 3G, IMT-Advanced, 4G (4th generation mobile communication system), 5G (5th generation mobile communication system), FRA (Future Radio Access), NR (New Radio), W-CDMA (registered trademark), GSM (registered trademark), CDMA2000, UMB (Ultra Mobile Broadband), IEEE 802.11 (Wi-Fi (registered trademark)), IEEE 802.16 (WiMAX (registered trademark)), IEEE 802.20, UWB (Ultra-Wideband), Bluetooth (registered trademark), or other suitable systems, and next-generation systems enhanced based on these. Furthermore, a combination of multiple systems (e.g., a combination of at least one of LTE and LTE-A with 5G) may also be applied.

[0075] The order of the procedures, sequences, flowcharts, etc. of each aspect / embodiment described in this disclosure may be changed unless it is consistent. For example, the methods described in this disclosure present elements of various steps using an example order, and are not limited to the particular order presented.

[0076] Input and output information may be stored in a specific location (for example, memory) or may be managed using a management table. Input and output information may be overwritten, updated, or added to. Output information may be deleted. Input information may be sent to another device.

[0077] The determination may be made based on a value represented by one bit (0 or 1), a Boolean value (true or false), or a numerical comparison (e.g., comparison with a predetermined value).

[0078] Although the present disclosure has been described in detail above, it is clear to those skilled in the art that the present disclosure is not limited to the embodiments described herein. The present disclosure can be implemented in modified and altered forms without departing from the spirit and scope of the present disclosure as defined by the claims. Therefore, the description of the present disclosure is intended to be illustrative and does not have any limiting meaning on the present disclosure.

[0079] Software, whether referred to as software, firmware, middleware, microcode, hardware description language, or otherwise, should be construed broadly to mean instructions, instruction sets, code, code segments, program code, programs, subprograms, software modules, applications, software applications, software packages, routines, subroutines, objects, executable files, threads of execution, procedures, functions, etc. Additionally, software, instructions, information, etc. may be transmitted or received over a transmission medium. For example, if software is transmitted from a website, server, or other remote source using wired technologies (such as coaxial cable, fiber optic cable, twisted pair, Digital Subscriber Line (DSL)), and / or wireless technologies (such as infrared, microwave), then such wired and / or wireless technologies are included within the definition of a transmission medium.

[0080] The information, signals, etc. described in this disclosure may be represented using any of a variety of different technologies. For example, data, instructions, commands, information, signals, bits, symbols, chips, etc. that may be referred to throughout the above description may be represented by voltage, current, electromagnetic waves, magnetic fields or magnetic particles, optical fields or photons, or any combination thereof. Note that terms described in this disclosure and terms necessary for understanding this disclosure may be replaced with terms having the same or similar meanings.

[0081] Furthermore, the information, parameters, etc. described in this disclosure may be expressed using absolute values, may be expressed using relative values ​​from a predetermined value, or may be expressed using other corresponding information.

[0082] As used in this disclosure, the phrase "based on" does not mean "based only on," unless expressly stated otherwise. In other words, the phrase "based on" means both "based only on" and "based at least on."

[0083] As used in this disclosure, any reference to an element using a designation such as "first," "second," etc. does not generally limit the quantity or order of those elements. These designations may be used in this disclosure as a convenient method of distinguishing between two or more elements. Thus, a reference to a first and a second element does not imply that only two elements may be employed or that the first element must in some way precede the second element.

[0084] The "unit" in the configuration of each of the above devices may be replaced with "means," "circuit," "device," or the like.

[0085] When the terms "include," "including," and variations thereof are used in this disclosure, these terms are intended to be inclusive, similar to the term "comprising." Furthermore, when the term "or" is used in this disclosure, it is not intended to be an exclusive or.

[0086] In this disclosure, where articles are added by translation, such as a, an, and the in English, the disclosure may include that the nouns following these articles are in the plural form.

[0087] In the present disclosure, the term "A and B are different" may mean "A and B are different from each other." The term may also mean "A and B are each different from C." Terms such as "separate" and "coupled" may also be interpreted in the same way as "different."

[0088] 1...information processing system, 9...network, 10...information processing device, 11...acquisition unit, 12...processing unit, 13...output unit, 17...learning unit, 18...storage unit, 19...control unit, 20...user terminal, 101...processor, 102...memory, 103...storage, 104...communication device

Claims

1. An information processing device having an acquisition unit that acquires feature values ​​of multiple items of multiple stores belonging to a target store group among one or more store groups consisting of multiple stores, and an output unit that outputs information regarding the peculiarity of unique items whose values ​​of store group feature values ​​are unique among the multiple items of store group feature values ​​calculated from the feature values ​​of the multiple stores belonging to the target store group.

2. The information processing device according to claim 1, wherein the information output by the output unit includes data on features of a plurality of stores belonging to the target store group, with peculiarities removed for the peculiar items.

3. An information processing device as described in claim 2, further comprising a processing unit that identifies, for each of a plurality of items of the store group feature in the target store group, the peculiar item using a condition that determines whether the item is the peculiar item, the condition relating to the distribution of values ​​of the store feature of a plurality of stores belonging to the target store group, and the output unit outputs data of the feature from which peculiarities have been removed from the feature determined by the processing unit to be the peculiar item.

4. The information processing device of claim 3, wherein when the target store group includes multiple store groups, the processing unit identifies the peculiar items using the condition that, among the features of the multiple store groups, there is an item for which a store group has a higher importance than other store groups, the peculiar items are the peculiar items.

5. The information processing device according to claim 3, wherein when the target store group consists of a single store group, the processing unit identifies the peculiar item using the condition that the item with the highest importance among the features of the single store group is the peculiar item.

6. An information processing device according to claim 2, further comprising a processing unit that performs processing to remove idiosyncrasies from the features of the plurality of stores in the idiosyncratic items, and wherein the output unit outputs data on the features from which the idiosyncrasies have been removed by the processing unit.

7. The information processing device according to claim 6, wherein the processing is a conversion processing for approximating the distribution of the feature data of the anomalous items to a normal distribution.

8. The information processing device according to claim 6, wherein the processing is processing for applying a weight vector to a vector of the feature quantities of the peculiar items.

9. The information processing device according to claim 2, further comprising a learning unit that uses the feature data output from the output unit to train a machine learning model for recommending stores to users.

10. An information processing method comprising the steps of: acquiring feature values ​​for multiple items of multiple stores belonging to a target store group among one or more store groups consisting of multiple stores; and outputting information regarding the uniqueness of a unique item, for which the value of the store group feature is unique, from the feature values ​​of the multiple stores in the unique item, calculated from the feature values ​​of the multiple stores belonging to the target store group.

Images