Information processing apparatus, information processing method, and program

Abstract

To enable an analysis of the user's action made to correspond to a consumption trend of the user.SOLUTION: An information processing apparatus includes: an acquisition section that acquires one or more pieces of flow line-related information pertaining to the movement of each of a plurality of users in an area and one or more pieces of purchase-related information pertaining to a plurality of pieces of purchase information when each of the plurality of users purchases a product in the area; an identification section that identifies a combination having high correlation among combinations of one or more pieces of flow line-related information and one or more pieces of purchase-related information acquired by the acquisition section; and an estimation section that estimates flow line-related information of a user from purchase-related information of the user on the basis of correlation relationship of the combination identified by the identification section.SELECTED DRAWING: Figure 9

Description

[Technical Field]

[0001] This invention relates to an information processing device, an information processing method, and a program. [Background technology]

[0002] Traditionally, stores such as supermarkets and shopping malls have been required to analyze information on how consumers move around within the facility in order to understand customer spending trends and further increase sales.

[0003] For example, Patent Document 1 proposes a location information collection device that collects customer user activity logs (movement history) on a shopping floor by receiving signals transmitted from a signal transmitter installed in a portable container that is portable and contains goods, which can be carried by the user within the shopping area. [Prior art documents] [Patent Documents]

[0004] [Patent Document 1] Japanese Patent Publication No. 2017-33442 [Overview of the project] [Problems that the invention aims to solve]

[0005] However, while the aforementioned technologies can track a customer's movement history within a store, they do not correlate this with purchase information (POS data) by store users. Therefore, it was not possible to analyze customer behavior in relation to their consumption trends. Furthermore, while it is possible to link customer behavior within the store with purchase information by introducing a system where customers purchase products by scanning product codes using their terminals, implementing such a system could incur significant costs. Therefore, the present invention aims to enable the analysis of user behavior in accordance with users' consumption trends. [Means for solving the problem]

[0006] One aspect of the present invention is an information processing device comprising: an acquisition unit that acquires one or more movement-related pieces of information relating to the movements of each of a plurality of users within an area, and one or more purchase-related pieces of information relating to a plurality of purchase information when each of the plurality of users purchases goods within the area; an identification unit that identifies a combination of the one or more movement-related pieces of information and the one or more purchase-related pieces of information acquired by the acquisition unit that has a high correlation; and an estimation unit that estimates the movement-related pieces of information of a user from the user's purchase-related information based on the correlation of the combination identified by the identification unit. [Effects of the Invention]

[0007] According to one aspect of the present invention, it becomes possible to analyze user behavior in accordance with users' consumption trends. [Brief explanation of the drawing]

[0008] [Figure 1] This is a schematic diagram illustrating the user behavior analysis system of the first embodiment. [Figure 2] This figure shows an example of the data structure of a movement pattern dataset. [Figure 3] This figure shows an example of the data structure of a POS dataset. [Figure 4] This figure illustrates movement-related data and purchase-related data in the user behavior analysis system of the first embodiment. [Figure 5] This figure shows an example of daily traffic flow and purchase data, as well as monthly traffic flow and purchase data. [Figure 6] This diagram illustrates the correlation coefficients between each data point related to customer movement patterns and each data point related to purchases. [Figure 7] This figure shows the results of regression analysis between purchase amount, stay duration, distance traveled, and number of customers. [Figure 8]It is a block diagram showing the internal configuration of each device of the user behavior analysis system according to the first embodiment. [Figure 9] It is a flowchart showing the operation of the user behavior analysis system according to the first embodiment. [Figure 10] It is a plan view showing an exemplary store floor in the user behavior analysis system according to the second embodiment. [Figure 11] It is a diagram showing an example of the data configuration of the product arrangement table. [Figure 12] It is a diagram showing the result of cluster analysis (principal component analysis). [Figure 13] It is a diagram exemplifying the flow line-related data and purchase-related data for each cluster. [Figure 14] It is a diagram showing the details of the flow line-related data and purchase-related data for each cluster. [Figure 15] It is a diagram showing the number of people staying in the zone for each cluster as a heat map. [Figure 16] It is a diagram showing the average stay time in the zone for each cluster as a heat map. [Figure 17] It is a diagram showing the average moving distance in the zone for each cluster as a heat map. [Figure 18] It is a diagram showing the categories of products purchased by users for each cluster in the order of purchase points. [Figure 19] It is a diagram showing the results of performing regression analysis on the purchase amount and each of the stay time, moving distance, and number of customers for a specific cluster. [Figure 20] It is a diagram showing the results of performing regression analysis on the purchase amount and each of the stay time, moving distance, and number of customers for a specific cluster. [Figure 21] It is a diagram showing the results of performing regression analysis on the purchase amount and each of the stay time, moving distance, and number of customers for a specific cluster. [Figure 22] It is a flowchart showing the operation of the user behavior analysis system according to the second embodiment.

Mode for Carrying Out the Invention

[0009] Hereinafter, an embodiment of the information processing apparatus, information processing method, and program of the present invention will be described. A first aspect of one aspect of the present invention is an information processing device comprising an acquisition unit, a specification unit, and an estimation unit. The acquisition unit acquires one or more movement-related information items related to the movements of each of the multiple users within the area, and one or more purchase-related information items related to multiple POS data (an example of purchase information) when each of the multiple users purchases products within the area. The POS data is aggregated by the POS system. The POS data includes information such as product codes that identify the purchased products, the number of items purchased, and the purchase amount. From the movement data, movement-related data such as the time users spend in the store and the distance they travel is acquired. From the POS data, purchase-related data such as the number of items purchased, the purchase amount, and the number of customers is acquired. The identification unit identifies combinations of one or more movement-related information and one or more purchase-related information acquired by the acquisition unit that have a high correlation. Here, "high correlation" means that an indicator showing the degree of correlation, such as the correlation coefficient, is higher than a predetermined threshold.

[0010] In this disclosure, "movement flow-related information" refers to information related to the movement of users within an area. As described later, the length of time users stay in an area and the distance users travel within an area are examples of movement flow-related information. In this disclosure, "purchase-related information" refers to information related to purchases made by users within the area. As described later, the number of items purchased, the purchase amount, and the average purchase price are examples of purchase-related information.

[0011] The estimation unit estimates user movement-related information from user purchase-related information based on the correlation of combinations identified by the identification unit. This allows the system to understand user behavior within the store from POS data of products purchased by users within the store. The following describes a user behavior analysis system that identifies user movement information from movement information of communication devices attached to product transport equipment such as shopping carts and shopping baskets, as an example of a system including an information processing device.

[0012] For example, in stores with multiple sales areas, such as supermarkets, many customers move around the store, pick up products from the sales areas, and pay at the cash register. Shopping baskets and carts are used to pick up products and carry them to the cash register. Therefore, in order to analyze the behavior of customers using the store, communication devices can be attached to shopping baskets and carts, and the movement of the shopping baskets and carts (i.e., the movement of the communication devices) can be acquired as movement data related to the customer's movements.

[0013] The following description assumes that the communication device attached to the shopping cart or shopping basket is a wireless tag. However, this does not limit the type of communication device to a wireless tag; it may also be a smartphone, wearable device, tablet, or other communication device used to acquire information about the user's movements.

[0014] A second aspect of one embodiment of the present invention further includes, in addition to the information processing device according to the first embodiment, an association unit that associates product acquisition information indicating whether or not each of multiple users acquired a product for each predetermined product placement area within an area with each of multiple movement data, and a cluster analysis unit that classifies the multiple movement data into multiple clusters based on the product acquisition information. In this case, the estimation unit estimates the movement information of a user from the user's purchase-related information for each of the multiple clusters classified by the analysis unit, based on the correlation of combinations identified by the identification unit. Since the behavioral characteristics of users within a store are considered to differ depending on the user's purpose of visiting the store, by clustering users based on the characteristics of their behavior within the store and estimating the user's movement information based on the correlation of each cluster, it is possible to estimate movement information with high accuracy according to the user's purpose of visiting the store.

[0015] In a third aspect of a certain embodiment of the present invention, in an information processing device according to the first or second embodiment, the identification unit excludes users whose purchase amount of goods exceeds a predetermined value and / or whose distance traveled within an area exceeds a predetermined value from among a plurality of users, and identifies highly correlated combinations based on one or more movement-related pieces of information and one or more purchase-related pieces of information of the remaining users. Users whose purchase amount of goods exceeds a predetermined value and / or whose distance traveled within an area exceeds a predetermined value are considered outliers because they deviate significantly from the average purchase amount and distance traveled per person, and excluding such users improves the overall accuracy of the analysis.

[0016] In a fourth aspect of a certain embodiment of the present invention, in any one of the first to third embodiments, the purchase-related information included in the highly correlated combination includes information indicating the purchase amount. In this case, the movement-related information included in the highly correlated combination includes information indicating the time users spend in the area, the distance users travel in the area, or the number of users. Since purchase amount, distance traveled, and the number of users are relatively easy to obtain, it is possible to construct a model more easily for estimating movement-related information.

[0017] In a fifth aspect of a certain embodiment of the present invention, in any one embodiment of the first to fourth aspects, the estimation unit performs regression analysis based on highly correlated combinations identified by the identification unit, generates an estimation model based on the regression analysis, and estimates movement-related information based on the generated estimation model. By using the estimation model based on regression analysis, movement-related information can be estimated with high accuracy.

[0018] A sixth aspect of one aspect of the present invention may further include a purchase trend analysis unit in addition to the information processing device according to the second aspect, which analyzes the purchase trends of users in each cluster based on the product categories of products purchased by users corresponding to movement information classified into each of a plurality of clusters. This makes it possible to associate the purchase trends of products of users corresponding to each cluster with movement-related information.

[0019] The average dwell time or average distance traveled by users in each product placement area, corresponding to the movement information classified into multiple clusters, may be visualized and displayed for each product placement area. This allows for the correlation between the behavioral characteristics of users within the area corresponding to each cluster and the movement-related information. Purchase-related information is the purchase amount, and movement-related information is the time spent in the area, the distance traveled in the area, and the number of users in the area. The estimation unit may estimate the time spent and distance traveled per person for a predetermined purchase amount based on the correlation. This makes it possible to quantitatively grasp, for example, the effect of changing the store layout. While the cluster analysis unit does not limit itself to a specific classification method, using the k-means algorithm is recommended. The k-means algorithm allows for efficient acquisition of classification results. Below, a user behavior analysis system according to one embodiment will be described in more detail with reference to the drawings.

[0020] (1) First Embodiment Figure 1 is a schematic diagram of the user behavior analysis system 1 of this embodiment. As shown in Figure 1, the user behavior analysis system 1 of this embodiment includes, for example, wireless tags 2 attached to carts CT or baskets used by each user in a store, a receiver 3, a store terminal 4, and a server 5 (an example of an information processing device). Figure 1 illustrates the case where the wireless tag 2 is attached to a cart CT.

[0021] Wireless tag 2 is an example of a communication device, such as a relatively small wireless communication device. Wireless tag 2 emits radio waves (beacon signals) at predetermined intervals. The beacon signals are emitted in such a way that they can be received within a predetermined range, depending on the surrounding radio wave environment. The beacon signals include, for example, identification information (such as a tag ID) that identifies wireless tag 2. Receiver 3 and server 5 are connected via a network NW, forming a location tracking system that identifies the location of a customer within the store floor (an example of an area). The network NW can be, for example, a cellular network, Wi-Fi network, the internet, a LAN (Local Area Network), a WAN (Wide Area Network), a public network, a dedicated network, or a wireless base station.

[0022] In one embodiment, the positioning of the wireless tag 2 utilizes the AOA (Angle of Arrival) method, in which a receiver 3 is installed on the ceiling of the store, the receiver 3 receives radio waves (beacon signals) emitted from the wireless tag 2, and calculates the incident angle of the received beacon signals. The receiver 3 measures the incident angle (direction of arrival) of the beacon signals received from the wireless tag 2 and sends the measured incident angle information to the server 5. The server 5 identifies the position (XY coordinates) of the wireless tag 2 from the position of the transmitting receiver 3 within the store (position in XYZ coordinates) and the incident angle based on that position. The communication protocol between the wireless tag 2 and the receiver 3 is not limited, but examples include Wi-Fi®, Bluetooth® Low Energy (BLE), etc.

[0023] While the location of the wireless tag 2 can be estimated using a single receiver 3 (locator), it is preferable to install more receivers 3 depending on the strength of the received signal strength (RSSI) of the beacon signal, the store area, and the radio wave environment of the store. For example, it is preferable to place receivers 3 at equal intervals on the ceiling of the store, and to place receivers 3 at shorter intervals in places where positioning accuracy is particularly required, such as in densely populated sales areas. Note that the positioning method for the wireless tag 2 is not limited to the AOA method, and other methods such as the TOA (Time of Arrival) method may also be used. The positioning interval for the wireless tag 2 can be set arbitrarily, but it should be set to the time necessary to accurately understand the user's actions (for example, 100ms to 2 seconds).

[0024] Server 5 calculates (acquires) the location of wireless tag 2 and records the movement data set for wireless tag 2. As shown in Figure 2, in the movement data set, time data (time information) and location data (location information) of wireless tag 2 are associated for each tag ID of wireless tag 2. In the following explanation, the data in which the time information and location information of wireless tag 2 are associated for a single tag ID is called "movement data" (an example of movement information). The movement data set contains multiple movement data corresponding to each of multiple tag IDs. The movement data corresponding to one tag ID corresponds to the movement of one user within the store.

[0025] Server 5 obtains a POS dataset from the POS system 7 that contains multiple POS data sets, each containing information such as the product code, quantity purchased, and purchase price of the product purchased by the user corresponding to the tag ID. Figure 3 shows an example of the data structure of the POS dataset. As shown in Figure 3, the POS dataset contains multiple POS data sets, each corresponding to a tag ID, and each POS data set contains the quantity purchased and the purchase price for each product code of the product purchased by the user corresponding to the tag ID. Server 5 associates movement data and POS data for each user's tag ID, retrieves one or more movement-related data points from the movement data, retrieves one or more purchase-related data points from the POS data, and identifies highly correlated combinations from the retrieved one or more movement-related data points and one or more purchase-related data points. Based on the identified highly correlated combinations, Server 5 generates an estimation model that, for example, estimates movement-related data from purchase-related data.

[0026] Referring again to Figure 1, the store terminal 4 is, for example, located in the store's office and equipped with a display panel such as a personal computer or tablet. The store terminal 4 can communicate with the server 5 via a network NW. The store terminal 4 is a terminal that, for example, acquires movement data from the server 5 and displays the movement patterns, or acquires and displays the execution results of a user behavior analysis program from the server 5, but it is not necessarily an essential component of the user behavior analysis system 1.

[0027] Next, referring to Figures 4 to 7, an example of how to construct an estimation model for estimating customer flow data from purchase-related data on Server 5 will be explained. Figures 4 to 7 are based on customer flow data and POS data collected over one month at an example store. In the following explanation, data for a single user, where customer flow data and POS data have been matched, may be referred to as a "sample." Matching means that customer flow data and POS data are associated on a one-to-one basis.

[0028] In the following explanation, "number of customers" refers to the number of users whose movement data and POS data have been matched. In other words, "number of users" in a fourth aspect of a certain aspect of the present invention corresponds to the number of customers. As shown in Figure 76, which will be referenced later, the correlation coefficient can be calculated using the number of customers (see Figure 6), and if a correlation is found, an estimation model can be generated as shown in Figure 7, which will be referenced later.

[0029] In one embodiment, one or more movement-related data points are extracted from the movement data, such as data indicating the time spent by customers in the store and the distance they traveled. One or more purchase-related data points are extracted from the POS data, such as data indicating the number of items purchased, the purchase amount, the average purchase price, and the number of customers. Figure 4 illustrates movement-related data and purchase-related data in the user behavior analysis system of the first embodiment. Figure 4 shows a macroscopic view of the results of aggregating movement-related data and POS data for a total of 3,472 users over one month at an exemplary store, illustrating movement-related data (time spent, distance traveled) and purchase-related data (number of items purchased, purchase amount, average purchase price). Figure 4 shows the mean, standard deviation, minimum value (min), 25th percentile (25%), 50th percentile (50%), 75th percentile (75%), and maximum value (max) for each of the following: duration of stay (seconds), distance traveled (m), number of items purchased (pieces), purchase amount (yen), and average purchase price (yen). As shown in Figure 4, the maximum values ​​(max) for distance traveled (m) and purchase amount (yen) are far from the mean and are considered outliers. It is preferable not to consider such outliers when conducting an overall analysis. In other words, it is preferable to exclude from the population users whose purchase amount exceeds a predetermined value, and / or whose distance traveled within the store exceeds a predetermined value. For example, it is advisable to apply a filter to exclude samples from the population that have a purchase amount exceeding 10,000 yen or a distance traveled exceeding 400m.

[0030] Figure 5 shows an example of daily traffic flow and purchase-related data, as well as an example of monthly traffic flow and purchase-related data. In Figure 5, Table 101 shows the daily values ​​for dwell time (seconds), distance traveled (m), number of items purchased (pieces), purchase amount (yen), and number of customers (people) over a one-month (30-day) period. Table 102 shows the mean, standard deviation (std), minimum value (min), 25th percentile (25%), 50th percentile (50%), 75th percentile (75%), and maximum value (max) for dwell time (seconds), distance traveled (m), number of items purchased (pieces), purchase amount (yen), and number of customers (people) for each day. In the example in Figure 5, the average purchase price is not aggregated, and the data used has been filtered to exclude samples with a purchase amount exceeding 10,000 yen and samples with a distance traveled exceeding 400m from the population.

[0031] Figure 6 illustrates the correlation coefficients between each data point in the movement-related data and each data point in the purchase-related data. In Figure 6, data is aggregated by date for the sample population that has undergone the above filtering process, and the correlations are shown for the survey items: time spent, distance traveled, number of items purchased, purchase amount, and number of customers. Figure 6 shows the correlation coefficients for each of the multiple combinations of movement-related data (one or more) and purchase-related data (one or more). Based on the correlation coefficients of each combination, it is possible to identify the combination with the highest correlation among the multiple combinations. In Figure 6, the combination with the highest correlation coefficient is shown with a thick line. Here, combinations with a correlation coefficient higher than a predetermined threshold can be selected as highly correlated combinations. The threshold can be set as appropriate, but for example, if the threshold is set to 0.9, then in the example in Figure 6, all combinations can be judged to have a high correlation. In other words, the purchase-related data included in highly correlated combinations includes information indicating the purchase amount and the number of items purchased, and the movement-related data included in highly correlated combinations includes information indicating the time users spend on the store floor, the distance users travel on the store floor, or the number of users (number of customers). In particular, among the one or more movement-related data, the correlation coefficient between distance traveled and purchase-related data is slightly higher than that between time spent on the store.

[0032] Figure 7 shows the results of regression analysis performed on combinations with high correlation, such as purchase amount and duration of stay, distance traveled, and number of customers. In the example in Figure 7, the regression analysis results (model) are shown when the dependent variables for purchase amount are duration of stay (seconds), distance traveled (m), and number of customers (people), respectively. In the example in Figure 7, for example, equation (1) is derived as a model where the dependent variable is the time spent, based on the purchase amount, and the coefficient of determination R 2 The figure was a high 0.936. y = 0.283x - 2549.47 (x: purchase amount (yen), y: duration of stay (seconds)) ... (1) According to equation (1), for example, when the purchase amount is 200,000 yen, the user's time spent in the store is predicted to be 54,093 seconds.

[0033] In the example in Figure 7, for example, equation (2) is derived as a model where the dependent variable is distance traveled, based on the purchase amount, and the coefficient of determination R 2 The result was a high 0.963. y = 0.068x - 360.708 (x: purchase price (yen), y: distance traveled (m)) ... (2) According to equation (2), for example, when the purchase amount is 200,000 yen, the distance a user travels within the store is predicted to be 13,287m.

[0034] In the example in Figure 7, for example, when the dependent variable is the number of customers, equation (3) is derived from the purchase amount, and the coefficient of determination R 2 The value was high at approximately 0.955. y = 0.00035372x + 21.256 (x: purchase amount (yen), y: number of customers (people)) ... (3) According to equation (3), for example, when the purchase amount is 200,000 yen, the number of customers is predicted to be 92.

[0035] Based on these results, for example, when the purchase amount is 200,000 yen, the estimated stay time is 54,093 seconds, the distance traveled is 13,287 m, and the number of customers is 92. Therefore, the average values ​​for the average stay time, distance traveled, and purchase amount per person are estimated to be 587.9 seconds, 144.4 meters, and 2,173 yen, respectively. When the average purchase amount per person is 1,000 yen, the estimated stay time and distance traveled are 270.1 seconds and 66.4 meters, respectively.

[0036] As shown in Figure 6 above, there is a high correlation between purchase amount and stay time, purchase amount and distance traveled, and purchase amount and number of customers in all combinations. Therefore, as shown in Figure 7, an estimation model can be constructed based on each combination. By using this estimation model, it is possible to estimate stay time, distance traveled, or number of customers from the purchase amount.

[0037] Next, the internal configuration of the user behavior analysis system 1 will be explained with reference to the block diagram in Figure 8. As shown in Figure 8, the wireless tag 2 includes, for example, a control unit 21 and a communication unit 22. The control unit 21 is mainly composed of a microcontroller and controls the entire wireless tag 2. For example, the control unit 21 performs processing on received signals and transmitted signals (processing of baseband signals). The communication unit 22 is an interface for communicating with the receiver 3, and for example modulates the transmission signal to the receiver 3 (e.g., a beacon signal) and broadcasts it according to BLE. The beacon signal includes the tag ID of the wireless tag 2.

[0038] As shown in Figure 8, the receiver 3 includes, for example, a radio wave receiving unit 31, an incident angle measuring unit 32, and a communication unit 33. The radio wave receiving unit 31 includes an antenna that receives beacon signals (radio waves) transmitted from the wireless tag 2. The incident angle measuring unit 32 measures the incident angle of the radio waves from the wireless tag 2 received by the radio wave receiving unit 31. The communication unit 33 is an interface for communicating with the wireless tag 2 and the server 5. For example, the communication unit 33 demodulates the received signal from the wireless tag 2. The communication unit 33 also associates the incident angle information measured by the incident angle measuring unit 32 with the tag ID included in the received beacon signal and transmits it to the server 5 via the network NW.

[0039] As shown in Figure 8, the store terminal 4 includes, for example, a control unit 41, a display unit 42, and a communication unit 43. The control unit 41 is mainly composed of a microcontroller and controls the entire store terminal 4. For example, the control unit 41 acquires movement data from the server 5 via the communication unit 43 for display and displays the movement data on the display unit 42, and / or acquires the execution results of a user behavior analysis program from the server 5 and displays them on the display unit 42. The display unit 42 includes, for example, a display panel such as an LCD (Liquid Crystal Display) panel and a drive circuit that drives the display panel based on display data acquired from the server 5. The communication unit 43 functions as a communication interface for communicating with the server 5 via a network NW.

[0040] As shown in Figure 8, the server 5 includes, for example, a control unit 51, storage 52, and a communication unit 53. The control unit 51 is mainly composed of a microcontroller and controls the entire server 5. For example, when the microcontroller of the control unit 51 executes a user behavior analysis program, the control unit 51 functions as an acquisition unit 511, a identification unit 512, an estimation unit 513, a mapping unit 514, and a cluster analysis unit 515. The mapping unit 514 and the cluster analysis unit 515 will be discussed in the second embodiment described later.

[0041] The acquisition unit 511 acquires one or more movement-related data points related to the movements of each of the multiple users on the store floor, and one or more movement-related data points related to the multiple POS data points when each of the multiple users purchases products on the store floor. The identification unit 512 identifies combinations with a high correlation among the combinations of one or more movement-related data and one or more purchase-related data acquired by the acquisition unit 511. In one embodiment, as described above, a filtering process is performed to exclude some samples from the population. That is, the identification unit 512 excludes users whose purchase amount of goods is above a predetermined value and / or users whose distance traveled within the store floor is above a predetermined value from among multiple users, and identifies combinations with a high correlation based on one or more movement-related data and one or more purchase-related data of the remaining users. If there are multiple combinations with a high correlation coefficient as shown in Figure 6 (for example, if there are multiple combinations with a correlation coefficient above a predetermined value), the identification unit 512 may preferentially identify the combinations with the highest correlation coefficients in order of their correlation coefficients. The estimation unit 513 estimates user movement-related data from user purchase-related data based on the correlation of combinations identified by the identification unit 512.

[0042] Storage 52 is a large-capacity storage device such as an HDD (Hard Disk Drive), and stores the movement data set (movement DS; see Figure 2) and the POS data set (POS DS; see Figure 3). Each piece of data in storage 52 can be updated, added, or deleted as appropriate in response to access from the control unit 51. The communication unit 53 functions as a communication interface for communicating with the receiver 3, store terminal 4, and POS system 7, etc., via the network NW.

[0043] Next, with reference to Figure 9, the process of calculating movement-related data performed by server 5 will be described. The movement-related data calculation process is performed by the control unit 51 of server 5 executing a user behavior analysis program. At the time the movement-related data calculation process is executed, it is assumed that the movement data of each individual user corresponding to the tag ID has been associated with the POS data.

[0044] Figure 9 is a flowchart showing the process for calculating movement-related data. First, the control unit 51 acquires movement data and POS data for individual users over a predetermined period (step S2). Each POS data includes information such as the product code that identifies the product purchased by the user, the number of items purchased, and the purchase amount. The data acquisition period (predetermined period) is not limited and may be one day, one week, or one month. Movement data is created for each tag ID based on data received from wireless tags 2 corresponding to each individual user. POS data is acquired from the POS system 7 in association with the tag ID. The control unit 51 acquires data (samples) by matching the movement data and POS data of each individual user via the tag ID. At this time, as shown in Figure 4, it is preferable to calculate the mean, standard deviation, minimum value, 25th percentile (25%), 50th percentile (50%), 75th percentile (75%), and maximum value (max) of, for example, the distance traveled (m) and purchase amount (yen), and set predetermined values ​​for filtering to exclude samples considered to be outliers from the population.

[0045] Next, the control unit 51 acquires one or more movement-related data (e.g., data on time spent and distance traveled) from the movement data and one or more purchase-related data (e.g., data on number of items purchased, purchase amount, and number of customers) from the POS data (step S4). The control unit 51 calculates a correlation coefficient between one of the acquired movement-related data and one of the acquired purchase-related data to identify combinations with high correlation. The control unit 51 generates an estimation model by performing regression analysis on the identified combinations (step S6). Next, the control unit 51 operates based on the generated estimation model. Specifically, for example, it obtains purchase-related data such as the purchase amount and the number of customers from the POS data obtained from the POS system 7, and inputs the obtained purchase amount and number of customers into the estimation model to obtain estimated values ​​for stay time and distance traveled (step S8).

[0046] As described above, according to the user behavior analysis system 1 of this embodiment, the server 5 acquires one or more movement-related data and one or more purchase-related data based on the movement data and POS data of each of the multiple users on the store floor. The server 5 finds combinations of data with a high correlation between the one or more movement-related data and the one or more purchase-related data, and uses an estimation model based on that combination to estimate the movement-related data of the user to be processed from the purchase-related data of that user. Therefore, it is possible to accurately grasp the behavior of a user from information on the purchase of products by that user on the store floor. Furthermore, according to the user behavior analysis system 1 of this embodiment, even if movement data is not available, movement-related data can be estimated from POS data by constructing an estimation model. Note that once estimation models have been created for multiple stores, the user behavior analysis system 1 may use the created estimation models to estimate movement-related data from POS data for other stores that are similar in location, store area, and customer attributes to any of the stores that were created for those stores.

[0047] In addition, the following effects can be obtained using the user behavior analysis system 1 of this embodiment. (I) For retailers with a large number of stores, implementing a system that matches and analyzes customer flow data and POS data based on the above-mentioned wireless tags would be expensive. However, costs can be reduced by building the above estimation model using only representative stores among the many stores and then horizontally deploying it to other stores. (II) If it is estimated that customers spend little time in the store or travel little distance within the store, measures can be considered to encourage impulse purchases by placing products that require little time spent in the store or travel little distance further inside the store, away from the entrance. (III) By placing product groups that tend to require customers to spend a long time in the store and travel long distances within the store in one location, it is possible to shorten the shopping time and distance customers travel within the store, and use the extra time to consider other products. (IV) It is conceivable to introduce a system in which customers purchase products by scanning product code information using a user terminal, thereby associating customer behavior within the store with information on products purchased by the customer. However, introducing such a system would incur significant costs. The user behavior analysis system 1 of this embodiment does not require the introduction of such a system, and can instead capture information on customer behavior within the store from, for example, product purchase information (POS data) by store users. Therefore, it is possible to capture customer behavior within the store from POS data of products purchased by customers within the store, and as described above, improve the store's product layout to increase sales and improve services for customers.

[0048] (2) Second embodiment The following describes the user behavior analysis system 1 of the second embodiment. For example, in stores such as drugstores and home centers, where the purpose of visiting differs for each individual user, generating an estimation model that takes into account the characteristics of each user improves the accuracy of estimating movement-related data, rather than generating an estimation model that estimates movement-related data using a sample of all users as the population. Therefore, in the user behavior analysis system 1 of this embodiment, users are clustered based on the characteristics of their behavior in the store, and an estimation model is generated for each cluster.

[0049] In one embodiment, the server 5 has a store map that includes two-dimensional information on the plan of the store floor where the user moves, and two-dimensional information regarding the boundary positions of each zone (an example of a product placement area) that divides the store floor. Figure 10 is a plan view of an exemplary store floor SF. Figure 10 shows an example in which the store floor SF is divided into 16 zones in a 4x4 grid, with zones Z11 to Z44. Although Figure 10 shows each zone as a rectangle or square, it is not limited to this, and each zone can be any shape depending on the overall floor shape and the products to be placed therein.

[0050] Server 5 has a product placement table that shows the product categories for products placed in each zone of the store floor. Figure 11 shows an example of the data structure of the product placement table. In the example shown in Figure 11, at least one of product categories C1 to C6 is set for each of zones Z11 to Z44. Here, product categories C1, C2, C3, C4, C5, and C6 are subdivided in this order, with category C1 being the broadest category and category C6 being the smallest category, representing the product itself. For example, in zone Z32, categories C1: processed foods, C2: curry, and C3: curry roux are set, and the categories are subdivided in this order. As shown in Figure 11, it is not necessary for all categories C1 to C6 to be set for each zone. Figure 11 shows an example where each zone is subdivided into six categories, but this is not the only option, and the number of categories when subdividing can be defined arbitrarily.

[0051] In the user behavior analysis system 1 of this embodiment, the microcontroller of the control unit 51 executes a user behavior analysis program, so that the control unit 51 functions as an acquisition unit 511, an identification unit 512, and an estimation unit 513, as well as an association unit 514 and a cluster analysis unit 515 (see Figure 8).

[0052] In the user behavior analysis system 1, for example, the product category values ​​corresponding to product codes included in POS data (see Figure 3) are predefined for each store. Therefore, based on the POS data, it is possible to identify the zone in which the corresponding user acquired the purchased product. The mapping unit 514 associates each piece of movement data with product acquisition information indicating whether or not each user acquired a product in a predetermined zone within the store floor. In other words, the product acquisition information includes information about the zone in which the user acquired the purchased product. The product acquisition information associated with each movement data can be represented by a 1x16 matrix, for example, if the store floor is divided into 16 zones as shown in Figure 10. For example, for movement data corresponding to the tag ID of wireless tag 2, product acquisition information can be represented by a matrix such as [0,1,0,0,…,0] or [0,0,0,1,…,1]. Each component of this matrix corresponds to one of the 16 zones Z11 to Z44 in Figure 10, and "1" indicates that the user purchased (acquired) a product in the corresponding zone, while "0" indicates that the user did not purchase (acquire) a product.

[0053] The cluster analysis unit 515 classifies multiple customer flow data into multiple clusters based on product acquisition information. Users with different reasons for visiting the store have their purchasing behavior reflected in their customer flow data. By binarizing the customer flow data and then performing cluster analysis, customer flow data and the users corresponding to that data can be appropriately classified. Specifically, the cluster analysis unit 515 applies machine learning clustering to product acquisition information, represented by a 1x16 matrix for each tag ID, to classify it into multiple clusters, thereby classifying the corresponding multiple movement data into multiple clusters. As a clustering method, for example, the k-means method can be used, but it is not limited to that, and other methods such as Ward's method can also be applied. The cluster analysis unit 515 classifies the movement data of customers with different reasons for visiting the store into multiple clusters.

[0054] The estimation unit 513 estimates user movement-related data from user purchase-related data for each of the multiple clusters analyzed by the cluster analysis unit 515, based on the correlation of combinations identified by the identification unit 512. For example, by constructing an estimation model targeting a sample population at the cluster level, the estimation accuracy of movement-related data can be improved for each group (cluster) of users with different reasons for visiting the store.

[0055] The following describes a specific example of the user behavior analysis system 1 of this embodiment. The specific example used here is the same store as in the first embodiment, and is the result of aggregating movement data and POS data corresponding to a total of 3,472 users over a one-month period. Figure 12 shows the results of cluster analysis (principal component analysis) using the k-means method based on product acquisition information obtained from each of the 3,472 movement data points. Here, it is shown that appropriate classification was achieved when the number of clusters was set to 3 (k=3). Note that the number of clusters is not limited to 3, and any value of 2 or 4 or more may be appropriate. The optimal number of clusters will vary depending on the sample population and should be determined appropriately for the store to which it is applied.

[0056] In the following explanation, cluster IDs 0, 1, and 2 are defined as identifying information for the three clusters. As shown in Figure 12, cluster ID 0 contains 650 movement data points, cluster ID 1 contains 1357 movement data points, and cluster ID 2 contains 1465 movement data points. Since each movement data point is associated with an individual user, this means that a total of 3472 users belong to one of the three clusters, cluster IDs 0 to 2.

[0057] Figure 13 is a diagram showing the average values ​​of each data point to provide an overview of movement-related data and purchase-related data for each cluster. The table in Figure 13 shows the average values ​​of movement-related data (time spent, distance traveled) and purchase-related data (number of items purchased, purchase amount, average purchase price) in stores, based on movement data and corresponding purchase data belonging to clusters with different cluster IDs. The bar graph in Figure 13 shows each average value shown in the table on the vertical axis. As shown in Figure 13, the cluster with cluster ID:0 has long time spent and distance traveled, and high values ​​for the number of items purchased, purchase amount, and average purchase price. The cluster with cluster ID:1 has fewer items purchased than other clusters, but moderate purchase amount and average purchase price. The cluster with cluster ID:2 has short time spent and distance traveled, low values ​​for purchase amount and average purchase price, but a large number of items purchased.

[0058] Figure 14 shows more detailed movement-related and purchase-related data for each cluster than Figure 13. In Figure 14, based on movement data belonging to clusters with different cluster IDs and corresponding purchase data, the mean, standard deviation, minimum value, 25th percentile (25%), 50th percentile (50%), 75th percentile (75%), and maximum value (max) are shown for each of the following: time spent (seconds), distance traveled (m), number of items purchased (pieces), purchase amount (yen), and average purchase price (yen). As shown in Figure 14, the cluster with cluster ID:0 has high percentile values ​​for all items. The cluster with cluster ID:1 has high values ​​for time spent, distance traveled, purchase amount, and average purchase price, but low values ​​for the number of items purchased. The cluster with cluster ID:2 has a high number of items purchased, but low values ​​for the other items.

[0059] From another perspective, heatmaps are used to understand the behavioral characteristics of users in each cluster. Figure 15 is a heatmap showing the number of people staying in each zone for each cluster in an example store. In the heatmap in Figure 15, darker shaded zones indicate that many users have passed through them. As shown in Figure 15, users belonging to cluster ID:0 move around the store floor evenly, excluding the children's zone and the nursing care products zone. Users belonging to cluster ID:1 pass through the children's zone and the nursing care products zone more often than other clusters, and pass through the cosmetics zone less often. Users belonging to cluster ID:2 mainly move around the food zone and pass through the daily necessities zone less often.

[0060] Figure 16 is a heatmap showing the average dwell time in each zone for each cluster in an example store. In the heatmap in Figure 16, darker shaded zones indicate longer average dwell times for users. As shown in Figure 16, users belonging to cluster ID:0 spend a long time in the daily necessities zone. The heatmaps corresponding to cluster ID:1 and cluster ID:2 are similar, and there is little difference in average dwell time for users in each cluster.

[0061] Figure 17 is a heatmap showing the average distance traveled in each zone for each cluster in an exemplary store. In the heatmap in Figure 17, darker shaded zones indicate longer average travel distances by users. As shown in Figure 17, users belonging to cluster ID:0 patrol the entire store floor, excluding the children's zone and the nursing care products zone. Users belonging to cluster ID:1 circulate around the central part of the store. Users belonging to cluster ID:2 do not go upwards from the food zone on the right side, but rather stay on the lower side.

[0062] From another perspective, in order to understand the purpose of customers' visits for each cluster, the trends in product categories purchased by customers in each cluster are analyzed. In this case, the control unit 51 functions as a purchase trend analysis unit that analyzes the purchase trends of customers in each cluster based on the product categories of products purchased by customers corresponding to the movement data classified into each of the multiple clusters. Figure 18 is a diagram showing the product categories purchased by customers in each cluster in order of the number of items purchased. As shown in Figure 18, customers belonging to cluster ID:0 often purchase products in the cosmetics category and can be called "beauty customers". Customers belonging to cluster ID:1 often purchase products in the laundry detergent category and can be called "daily necessities customers". Customers belonging to cluster ID:2 often purchase products in the beverages and alcoholic beverages category and can be called "food customers". By analyzing the purchase trends of customers in each cluster, it is possible to correlate the purchase trends of customers in each cluster with the movement-related data obtained by the estimation model described later.

[0063] Figures 19 to 21 show the results of regression analysis performed on purchase amount, stay time, distance traveled, and number of customers, similar to Figure 7. Figures 19 to 21 represent populations of users belonging to clusters ID: 0 to 2, respectively. In Figure 19, for users belonging to cluster ID: 0, for example, equation (4) is derived as a model where the dependent variable is stay time, derived from purchase amount, and the coefficient of determination R 2 The value was a high 0.925. y = 0.230x - 1587.977 (x: purchase amount (yen), y: duration of stay (seconds)) ... (4) According to equation (4), for example, when the purchase amount is 200,000 yen, the user's time spent in the store is predicted to be 44,419 seconds.

[0064] In Figure 19, for users belonging to cluster ID:0, equation (5) is derived as a model where the dependent variable is distance traveled, for example, from the purchase amount, and the coefficient of determination R 2 The figure was a high 0.943. y = 0.056x - 314.644 (x: purchase price (yen), y: distance traveled (m)) ... (5) According to equation (5), for example, when the purchase amount is 200,000 yen, the distance traveled by the user within the store is predicted to be 10,894m.

[0065] In Figure 19, for users belonging to cluster ID:0, equation (6) is derived as a model where the dependent variable is the number of customers, for example, from the purchase amount, and the coefficient of determination R 2 The result was a high 0.938. y = 0.00024928x + 2.144 (x: purchase amount (yen), y: number of customers (people)) ... (6) According to equation (6), for example, when the purchase amount is 200,000 yen, the number of customers is predicted to be 52.

[0066] Based on the results shown in Figure 19, for users belonging to cluster ID:0 (beauty-related customers), if the purchase amount is 200,000 yen, the estimated stay time is 44,419 seconds, the distance traveled is 10,894 m, and the number of customers is 52. Therefore, the average values ​​for the stay time, distance traveled, and purchase amount per person are estimated to be 854.2 seconds, 209.5 meters, and 3,846 yen, respectively. When the purchase amount per person is 1,000 yen, the estimated stay time and distance traveled are 222 seconds and 54.4 meters, respectively.

[0067] In Fig. 20, for the users belonging to Cluster ID: 1, when the target variable is the stay time from the purchase amount, for example, Equation (7) is derived as the model, and the coefficient of determination R 2 is as high as about 0.872. y = 0.281x + 947.520 (x: purchase amount (yen), y: stay time (seconds))... (7) According to Equation (7), when the purchase amount is 200,000 yen, the stay time of the user in the store is predicted to be 57,203 seconds.

[0068] In Fig. 20, for the users belonging to Cluster ID: 1, when the target variable is the moving distance from the purchase amount, for example, Equation (8) is derived as the model, and the coefficient of determination R 2 is as high as about 0.943. y = 0.069x + 234.643 (x: purchase amount (yen), y: moving distance (m))... (8) According to Equation (8), for example, when the purchase amount is 200,000 yen, the moving distance of the user in the store is predicted to be 14,077 m.

[0069] In Fig. 20, for the users belonging to Cluster ID: 1, when the target variable is the number of customers from the purchase amount, for example, Equation (9) is derived as the model, and the coefficient of determination R 2 is as high as about 0.919. y = 0.0003735x + 7.301 (x: purchase amount (yen), y: number of customers (persons))... (9) According to Equation (9), for example, when the purchase amount is 200,000 yen, the number of customers is predicted to be 82 persons.

[0070] From the results shown in Fig. 20, for the users (general merchandise customers) belonging to Cluster ID: 1, when the purchase amount is 200,000 yen, for example, the stay time is predicted to be 57,203 seconds, the moving distance is 14,077 m, and the number of customers is 82 persons. Therefore, the average values of the stay time per person, the moving distance per person, and the purchase amount per person are estimated to be 697.5 (seconds), 171.6 (m), and 2,439 (yen), respectively. When the purchase amount per person is 1,000 yen, the stay time and the moving distance are estimated to be 286 (seconds) and 70.3 (m), respectively.

[0071] In Figure 21, for users belonging to cluster ID:2, equation (10) is derived as a model where the dependent variable is the time spent, for example, from the purchase amount, and the coefficient of determination R 2 The value was high at approximately 0.845. y = 0.288x + 1799.359 (x: purchase amount (yen), y: duration of stay (seconds)) ... (10) According to equation (10), for example, when the purchase amount is 200,000 yen, the user's time spent in the store is predicted to be 59,311 seconds.

[0072] In Figure 21, for users belonging to cluster ID:2, equation (11) is derived as a model where, for example, the dependent variable is distance traveled from the purchase amount, and the coefficient of determination R 2 The result was a high 0.899. y = 0.069x + 475.910 (x: purchase price (yen), y: distance traveled (m)) ... (11) According to equation (11), for example, when the purchase amount is 200,000 yen, the distance traveled by the user within the store is predicted to be 14,213m.

[0073] In Figure 21, for users belonging to cluster ID:2, equation (12) is derived as a model where the dependent variable is the number of customers, for example, from the purchase amount, and the coefficient of determination R 2 The result was a high 0.891. y = 0.00043008x + 10.984 (x: purchase amount (yen), y: number of customers (people)) ... (12) According to equation (12), for example, when the purchase amount is 200,000 yen, the number of customers is predicted to be 97.

[0074] Based on the results shown in Figure 21, for users belonging to cluster ID:2 (food-related customers), if the purchase amount is 200,000 yen, the estimated stay time is 59,311 seconds, the distance traveled is 14,213 m, and the number of customers is 97. Therefore, the average values ​​for the stay time, distance traveled, and purchase amount per person are estimated to be 611.4 (seconds) (=59,311 / 97), 146.5 (m) (=14,213 / 97), and 2,061 (yen) (=200,000 / 97), respectively. When the purchase amount per person is 1,000 yen, the estimated stay time and distance traveled are 296.5 (seconds) (=59,311 × 1000 / 200,000) and 71 (m) (=14,213 × 1000 / 200,000), respectively.

[0075] As described above, in this embodiment, as shown in Figure 21, an estimation model is constructed individually for each of the multiple clusters that take into account the behavioral characteristics of users within the store. In other words, estimation models are constructed that capture the characteristics of customers with different objectives. By using the models constructed in this way, customer movement-related data can be estimated with high accuracy according to the characteristics of the customer.

[0076] Next, the operation of the user behavior analysis system 1 of this embodiment will be described with reference to Figure 22. Figure 22 is a flowchart of the movement-related data calculation process of this embodiment. The flowchart shown in Figure 22 differs from that in Figure 9 in that step S3, in which cluster analysis is performed, has been added. In cluster analysis, the control unit 51 acquires product acquisition information corresponding to each movement data based on the movement data and POS data of individual users for a predetermined period acquired in step S2. The product acquisition information indicates whether or not each of multiple users acquired a product in a predetermined zone within the store floor. As mentioned above, the product acquisition information can be represented as a matrix (with components of 0 or 1) corresponding to the number of zones.

[0077] In cluster analysis, the control unit 51 performs machine learning clustering on the product acquisition information corresponding to each of the multiple movement data sets, and classifies them into multiple clusters. These multiple clusters group users based on their purchasing behavior, and it is assumed that users belonging to each cluster have different reasons for visiting the store. Steps S4 to S8 in Figure 22 are the same as the corresponding processes in the flowchart shown in Figure 9, except that they are performed on a cluster basis. In the user behavior analysis system 1 of this embodiment, for each cluster, combinations of data with a high correlation between one or more purchase-related data and one or more movement-related data, which are the target variables, are identified, and an estimation model is generated by performing regression analysis based on the identified combinations. Since an estimation model is generated for each cluster, highly accurate estimates of dwell time and travel distance can be obtained according to the user's purpose of visiting the store.

[0078] Based on estimated customer flow data tailored to the characteristics of each customer cluster, stores can consider promotional strategies. For example, if it is predicted that customers who frequently purchase a particular product tend to have short dwell times and short distances traveled, the store can change the product layout to increase these dwell times and distances, thereby encouraging impulse purchases. Alternatively, if customers who frequently purchase a particular product tend to have long dwell times and distances traveled, the store can consolidate those products in one location to shorten shopping time and distance, allowing customers to use that time to consider other purchases.

[0079] In the embodiments described above, the case in which data is exchanged between the store terminal 4 and the server 5 is described as being performed via a network NW, but this is not the only case. Data can also be exchanged between the store terminal 4 and the server 5 via storage media such as a USB (Universal Serial Bus) memory, an SD (Secure Digital) memory card, an HDD device, or an SSD (Solid State Drive). The receiver 3 and server 5 are not limited to communicating via a network NW; they may also communicate one-to-one via wired or wireless connections. The functions of server 5 may be distributed across multiple devices, or the storage 52 of server 5 may be distributed across multiple devices.

[0080] Although embodiments of the information processing device, information processing method, and program have been described above, the present invention is not limited to the above embodiments. Furthermore, the above embodiments can be improved or modified in various ways without departing from the spirit of the present invention. For example, equations (1) to (12) in the first and second embodiments are examples, and the output (y) corresponding to the input (x) may be derived using other functions, tables, trained models, etc. [Explanation of Symbols]

[0081] 1…User behavior analysis system 2… Wireless tags 21... Control Unit 22… Communications Department 3…Receiver 31...Radio wave receiving unit 32...Incidence angle measurement section 33… Communications Department 4…Store terminals 41... Control Unit 42...Display section 43... Communications Department 5…Server 51... Control Unit 511…Acquisition Department 512…Specific part 513... Estimation section 514...Matching section 515...Cluster Analysis Department 52...Storage 53... Communications Department 7…POS system 101...Movement flow data 102…Behavioral dataset NW...Network CT... Cart SF...Store floor Z11~Z44... Zone

Claims

1. an acquisition unit that acquires one or more pieces of flow-line-related information related to the movement of each of a plurality of users within an area and one or more pieces of purchase-related information related to a plurality of purchase information when each of the plurality of users purchases a product within the area; an identification unit that identifies a combination having a high correlation from among combinations of one or more pieces of movement-related information and one or more pieces of purchase-related information acquired by the acquisition unit; an estimation unit that estimates movement line related information of the user from the purchase related information of the user based on the correlation of the combination identified by the identification unit; An information processing device comprising:

2. a matching unit that matches each of the plurality of pieces of movement line information with product acquisition information indicating whether each of the plurality of users has acquired a product for each predetermined product placement area within the area; a cluster analysis unit that classifies the plurality of pieces of movement line information into a plurality of clusters based on the product acquisition information, the estimation unit estimates movement line related information of the user from the purchase related information of the user based on the correlation of the combinations identified by the identification unit for each of the plurality of clusters classified by the cluster analysis unit; 2. An information processing device according to claim 1.

3. the identification unit excludes, from among the plurality of users, users whose product purchase amount is equal to or exceeds a predetermined value and / or users whose travel distance within the area is equal to or exceeds a predetermined value, and identifies the combination having a high correlation based on the one or more pieces of movement line related information and the one or more pieces of purchase related information of the remaining users.

3. An information processing device according to claim 1 or 2.

4. the purchase-related information included in the highly correlated combination includes information indicating a purchase amount, The flow line related information included in the combination with high correlation includes information indicating a stay time of the user in the area, a moving distance of the user in the area, or a number of the multiple users.

3. An information processing device according to claim 1 or 2.

5. the estimation unit performs a regression analysis based on the highly correlated combinations identified by the identification unit, generates an estimation model based on the regression analysis, and estimates traffic line related information based on the generated estimation model.

3. An information processing device according to claim 1 or 2.

6. a purchase tendency analysis unit that analyzes the purchase tendency of users in each cluster based on a product category of a product purchased by a user corresponding to the movement line information classified into each of the plurality of clusters, 3. An information processing device according to claim 2.

7. acquiring one or more pieces of flow-line-related information related to the flow lines and one or more pieces of purchase-related information related to the purchases based on information in which a plurality of pieces of flow-line information related to the movements of each of a plurality of users within an area is associated with a plurality of pieces of purchase information when each of the plurality of users purchases products within the area; identifying a combination having a high correlation from among the combinations of one or more pieces of movement-related information and one or more pieces of purchase-related information acquired in the acquiring step; a step of estimating movement line related information of the user from the purchase related information of the user based on the correlation of the combinations identified by the identifying step; An information processing method comprising:

8. A program for causing a computer to execute a predetermined method, The method comprises: acquiring one or more pieces of flow-line-related information related to the flow lines and one or more pieces of purchase-related information related to the purchases based on information in which a plurality of pieces of flow-line information related to the movements of each of a plurality of users within an area is associated with a plurality of pieces of purchase information when each of the plurality of users purchases products within the area; identifying a combination having a high correlation from among the combinations of one or more pieces of movement-related information and one or more pieces of purchase-related information acquired in the acquiring step; and estimating movement line related information of the user from the purchase related information of the user based on the correlation of the combinations identified by the identifying step. program.

Images