JP7715328B2 - Information processing method, program, storage medium, and information processing device - Google Patents

Description

The present invention relates to an information processing method, a program, a storage medium , and an information processing device.

Traditionally, behavioral history is extracted from data containing human behavior and then analyzed. The results of analyzing behavioral history are used for a variety of purposes, such as product and service marketing, recommendations, and transportation planning.

As an example of a method for analyzing the above-mentioned behavioral history, Patent Document 1 discloses a method for analyzing user behavioral trends based on the usage history of a contactless transportation IC card or a mobile device with equivalent functionality. Specifically, based on the usage history, usage trends such as frequency of station use, length of stay at the station, and whether stores visited within or around the station are entertainment-oriented or convenient are calculated, and station users are characterized and classified using these index values. Then, based on the user classification results, information such as recommendations targeted at purchasing behavior within or around the station is provided.

JP 2011-76291 A

After extensive research, the inventors discovered that classifying a series of behavioral histories into predetermined behavioral transition patterns is useful for achieving the various objectives mentioned above. However, conventional behavioral analysis methods, such as the method disclosed in Reference 1, do not take such classification into consideration, and there is room for improvement.

The present invention was made in light of these issues and aims to appropriately analyze a person's behavioral history.

In order to achieve the above-mentioned object, the information processing method of the present invention is characterized in that an information processing device executes the following steps: an extraction step in which, from data including human behavior, an extraction of a temporally continuous behavioral history from the start to the end of the behavior for each identifier that identifies an individual; a mapping step in which the behavioral history is vectorized and mapped into a vector space ; a calculation step in which, using the mapped behavioral history, the temporal proximity between the first and last behaviors in the behavioral history and/or the number of stays in the behavioral history are calculated; and a classification step in which the behavioral history is classified into behavioral transition patterns based on the proximity and/or the number of stays .
In addition, the information processing method of the present invention is characterized in that an information processing device executes the following steps: an extraction step in which, from data including human behavior, an extraction history of a temporally continuous behavior from the start to the end of the behavior for each identifier that identifies an individual; a mapping step in which the behavior history is mapped to a geographical space; a calculation step in which, using the mapped behavior history, the proximity of the first behavior to the last behavior in terms of time in the behavior history and/or the number of stays in the behavior history are calculated; and a classification step in which the behavior history is classified into behavior transition patterns based on the proximity and/or the number of stays.

In the information processing method, the calculation step may calculate both the proximity and the number of stays, and the classification step may classify the behavior history into behavior transition patterns by shape based on the proximity and the number of stays. Note that the shape is formed by mapping the behavior history onto a geographic space or vector space and connecting the mapped behavior history with lines.

In the information processing method, the calculation step may calculate the proximity as the physical distance between the locations where the first action and the last action were performed. Alternatively, the calculation step may calculate the proximity as the similarity between the first action and the last action.

In the information processing method, the calculation step may include a step of calculating the proximity of two temporally consecutive actions and/or the time between the two actions, a step of determining whether the two actions are a stay or a movement based on the proximity of the two actions and/or the time between the two actions, and a step of calculating the number of stays. Alternatively, in the calculation step, the number of actions in the action history may be calculated as the number of stays.

In the information processing method, the data may include identification information of the person who performed the action.

In the information processing method, the data may include data obtained from a transportation IC card or a mobile terminal.

According to another aspect of the present invention, there is provided a program for causing an information processing device to execute the information processing method.

According to yet another aspect of the present invention, there is provided a computer-readable storage medium storing the program.

From another perspective, the present invention is an information processing device comprising: a memory unit that stores a program for controlling the information processing device; a calculation unit that controls the information processing device in accordance with the program ; an extraction unit that extracts, from data including human behavior, a temporally continuous behavioral history from the start to the end of the behavior for each identifier that identifies an individual; a mapping unit that vectorizes the behavioral history and maps it into a vector space; a calculation unit that uses the mapped behavioral history to calculate one or both of the proximity between the first behavior and the last behavior in the behavioral history and the number of stays in the behavioral history; and a classification unit that classifies the behavioral history into behavior transition patterns based on one or both of the proximity and the number of stays.
The present invention is also characterized by an information processing device comprising: a memory unit that stores a program for controlling the information processing device; a calculation unit that controls the information processing device in accordance with the program; an extraction unit that extracts, from data including human behavior, a temporally continuous behavioral history from the start to the end of the behavior for each identifier that identifies an individual; a mapping unit that maps the behavioral history into geographic space; a calculation unit that uses the mapped behavioral history to calculate one or both of the proximity between the first behavior and the last behavior in the behavioral history and the number of stays in the behavioral history; and a classification unit that classifies the behavioral history into behavior transition patterns based on one or both of the proximity and the number of stays.

In the information processing device, the calculation unit may calculate both the proximity and the number of stays, and the classification unit may classify the behavior history into behavior transition patterns by shape based on the proximity and the number of stays. Note that the shape is formed by mapping the behavior history onto a geographic space or vector space and connecting the mapped behavior histories with lines.

In the information processing device, the calculation unit may calculate the proximity as the physical distance between the locations where the first action and the last action were performed. Alternatively, the calculation unit may calculate the proximity as the similarity between the first action and the last action.

In the information processing device, the calculation unit may calculate one or both of the proximity between two consecutive actions in time and the time between the two actions, and determine whether the two actions are a stay or a movement based on one or both of the proximity between the two actions and the time between the two actions, and calculate the number of stays. Alternatively, the calculation unit may calculate the number of actions in the action history as the number of stays.

In the information processing device, the data may include identification information of the person who performed the action.

In the information processing device, the data may include data acquired from a transportation IC card or a mobile terminal.

This invention classifies a person's behavioral history into behavioral transition patterns, allowing for appropriate analysis of that behavioral history. As a result, the analysis results can be used for a variety of purposes, such as product and service marketing and recommendations, event planning, and transportation planning.

1 is an explanatory diagram showing an outline of the configuration of an information processing system according to an embodiment of the present invention; FIG. 2 is a flowchart showing main steps of an information processing method according to the first embodiment. 4 is a table illustrating an example of behavior history data in the first embodiment. 10 is a table showing master data of stations in the first embodiment. 10 is a table showing the results of extracting behavior history from behavior history data in the first embodiment. 10 is a table showing the results of mapping behavioral histories in the first embodiment. 10 is a table showing the results of calculating the closeness between two actions in the first embodiment. 10 is a table showing the results of calculating start-end behavior proximity in the first embodiment. 10 is a table showing the results of classifying behavior histories into behavior transition patterns in the first embodiment. 1 is a graph that visualizes a behavior transition pattern in the first embodiment. FIG. 10 is a flowchart showing main steps of an information processing method according to a second embodiment. 10 is a table showing the results of calculating the number of stays in the second embodiment. 10 is a table showing the results of classifying behavior histories into behavior transition patterns in the second embodiment. 10 is a graph visualizing a behavior transition pattern in the second embodiment. FIG. 10 is a flowchart showing main steps of an information processing method according to a third embodiment. 13 is a table showing the results of deriving the station where the maximum elapsed time was spent in the third embodiment. 13 is a graph visualizing the results of the aggregation of the number of passengers by station in the third embodiment. FIG. 10 is a flowchart showing main steps of an information processing method according to a fourth embodiment. 13 is a table showing an example of behavior history data in the fourth embodiment. 13 is a table showing master data of a store in the fourth embodiment. 13 is a table showing master data of items in the fourth embodiment. 13 is a table showing the results of extracting behavior history from behavior history data in the fourth embodiment. 13 is a table showing the results of vectorizing a behavior history in the fourth embodiment. 13 is a table showing the results of extracting the first action and the last action in the action history in the fourth embodiment. 13 is a table showing the results of calculating start-end behavior proximity in the fourth embodiment. 13 is a table showing the results of calculating the number of stays in the fourth embodiment. 13 is a table showing classification rules for behavior transition patterns in the fourth embodiment. 13 is a table showing the results of classifying behavior histories into behavior transition patterns in the fourth embodiment. 13 is a graph visualizing a behavior transition pattern in the fourth embodiment.

Embodiments of the present invention will be described below with reference to the drawings. Note that in this specification and drawings, components having substantially the same functional configuration will be assigned the same reference numerals, and redundant description will be omitted.

<Configuration of Information Processing System 1>
FIG. 1 is an explanatory diagram showing an outline of the configuration of an information processing system 1 according to this embodiment.

The information processing system 1 has a database device 10 and an information processing device 20. The database device 10 and the information processing device 20 are configured to be connectable via a network 30. The network 30 is not particularly limited as long as it allows communication, but examples of the network include the Internet, a wired LAN, and a wireless LAN. Note that in this embodiment, only one database device 10 is shown to simplify the explanation, but the information processing system 1 includes multiple database devices 10.

(Configuration of database device 10)
The database device 10 includes a calculation unit 11 , a storage unit 12 , and a communication unit 13 .

The calculation unit 11 is a circuit (hardware) or a central processing unit such as a CPU. The calculation unit 11 controls the database device 10 in accordance with a program (software) stored in the memory unit 12.

The memory unit 12 stores a database of data to be output to the information processing device 20. This data includes data containing people's behavior (hereinafter sometimes referred to as "behavior history data"). In addition to people's behavior, the behavior history data also includes identification information of the people.

A person's identification information is an identifier that can identify one individual from another, and includes, for example, a user ID. Note that the identifier may also be a combination of other factors in addition to the user ID, such as date, time, or location. For example, Person A on January 1st and Person A on January 2nd may be treated as different individuals and assigned different identifiers. A person's identification information may also include attribute information linked to the individual (for example, age, gender, etc.).

In other words, behavioral history data is data (transaction data) that includes actions that an individual intentionally performs, and is linked to that individual. Furthermore, behavioral history data includes actions represented by points on a time axis, and a series of these temporally consecutive actions constitutes a behavioral history. Note that the type of behavioral history data is not particularly limited, and includes a variety of data.

The behavioral history data includes, for example, information acquired from a transportation IC card or a mobile device with equivalent functionality (hereinafter collectively referred to as a "transportation IC card"). For example, when a transportation IC card is used at a ticket gate at a train station or on a bus or other means of transportation, the usage history is sent to the database device 10 and stored in the memory unit 12. Similarly, when a transportation IC card is used at a card reading terminal, POS register terminal, or vending machine installed in a store outside of a transportation facility, such as inside or around a station, the usage history is sent to the database device 10 and stored in the memory unit 12. When the usage history is stored in this way, the user ID that identifies the transportation IC card and the location information of the location where the transportation IC card was used are also linked to the usage history and stored.

The behavioral history data also includes, for example, website operation history. Website operation history includes, for example, product purchase history and service usage history on an EC (electronic commerce) site, and website browsing history. When this website operation history is stored in the memory unit 12, the user ID for the website is also stored in association with the operation history.

Furthermore, behavioral history data may include, for example, purchase data (POS data) at retail stores, credit card and electronic money usage data, transaction data at financial institutions (e.g., deposits, withdrawals, payments, etc.), food and drink data at restaurants, accommodation data at accommodation facilities, playback and play data for music and video services and games, and point usage data at point sites.

The memory unit 12 also stores various programs for controlling the database device 10.

The communication unit 13 is a communication interface that mediates communication with the network 30 and performs data communication with the information processing device 20.

(Configuration of information processing device 20)
The information processing device 20 includes a calculation unit 21 , a storage unit 22 , an extraction unit 23 , a mapping unit 24 , a calculation unit 25 , a classification unit 26 , a visualization unit 27 , and a communication unit 28 .

The calculation unit 21 is a circuit (hardware) or a central processing unit such as a CPU. The calculation unit 21 controls the information processing device 20 in accordance with a program (software) stored in the memory unit 22.

The memory unit 22 stores various data processed by the information processing device 20. For example, the memory unit 22 stores input data input to the information processing device 20 from the database device 10, data resulting from extraction by the extraction unit 23, data resulting from mapping by the mapping unit 24, data resulting from calculation by the calculation unit 25, data resulting from classification by the classification unit 26, and data visualized by the visualization unit 27. The memory unit 22 also stores various programs for controlling the information processing device 20.

The extraction unit 23 extracts behavioral history from the behavioral history data input from the database device 10 to the information processing device 20. As described above, behavioral history is linked to individuals, and the extraction unit 23 extracts behavioral history for each identifier that identifies an individual. The extracted behavioral history is a temporally continuous behavioral history from the start to the end of the behavior. The time from the start to the end of the behavior can be set arbitrarily. For example, the time may be one day, or, if the person's behavior is entering and exiting a station, the time may be the time from when the station opens to when it closes. Furthermore, if the person's behavior is recorded, the time may be the time from the start record to the end record. Furthermore, if the person's behavior is a product purchase history or service usage history on an e-commerce site, the time may be the time from when they log in to when they log out of the e-commerce site. Furthermore, the time may be the time from when they wake up to when they go to bed, or the time from when they leave home to when they return home. Note that the raw data input from the database device 10 may contain unnecessary information, and the extraction unit 23 also performs a cleansing process to remove such unnecessary information.

The mapping unit 24 maps the behavioral history extracted by the extraction unit 23. Specifically, the mapping unit 24 vectorizes the behavioral history and maps it into a vector space. When vectorizing the behavioral history, parameters necessary for processing are selected from the multiple parameters included in each behavior, and each behavior is vectorized. The parameter selection method is arbitrary; for example, the parameters may be selected by an operator or by machine learning. The vector space may be a vector representation of a real space that physically exists, or a vector representation of a virtual space that does not physically exist. For example, if the behavioral history to be processed is behavior on public transportation, in a brick-and-mortar store, etc., the vector space will be a vector representation of the real space. On the other hand, if the behavioral history to be processed is behavior on a website, etc., the vector space will be a vector representation of the virtual space.

The parameter to be vectorized for each action may be a single parameter, or a parameter obtained by reducing the dimension of multiple parameters. When reducing the dimension of multiple parameters, the multiple parameters are input into a matrix, and principal components are extracted, for example, by performing principal component analysis, to generate a parameter with reduced dimension.

Furthermore, if the behavioral history data can be linked to geographic information, the mapping unit 24 may map the behavioral history into geographic space.

The calculation unit 25 calculates the temporal proximity between the first and last actions in the action history (hereinafter sometimes referred to as "start-end action proximity") and/or the number of stays in the action history.

When calculating the proximity between two actions, for example, if the action history to be processed is an action on public transportation, in a brick-and-mortar store, or the like, and can be linked to geographic information, the calculation unit 25 calculates the physical distance between the locations where the two actions were performed as the proximity. The physical distance may be, for example, a geographical straight-line distance or a distance along a geographical route.

Furthermore, when calculating the closeness between two actions, for example, if the action history to be processed is an action on a website or the like and the action history can be vectorized, the calculation unit 25 calculates the similarity between the two actions as the closeness. The similarity can be any measure, but examples include cosine similarity, Euclidean distance, and Manhattan distance.

When calculating the number of stays in the behavior history, the calculation unit 25 performs, for example, the following steps (1) to (3).
(1) Calculate the proximity of two temporally consecutive actions and/or the time between the two actions. Hereinafter, two temporally consecutive actions will be simply referred to as "two actions," and the proximity of the two actions will be referred to as "consecutive action proximity."
(2) Based on one or both of the consecutive behavioral proximity calculated in (1) above and the time between the two behaviors (hereinafter sometimes referred to as “consecutive behavioral time”), the two behaviors are determined as a stay or a move.
(3) The number of stays determined in (2) above is calculated as the number of stays.

In (1) above, the method for calculating consecutive behavior proximity is the same as the method for calculating start-end behavior proximity. That is, the calculation unit 25 may calculate the physical distance between the locations where two behaviors were performed as consecutive behavior proximity, or may calculate the similarity between the two behaviors as consecutive behavior proximity.

In (2) above, when determining whether a period between two actions is a stay or a move, the period between the two actions includes the state of a person between two actions that are consecutive in time. A stay indicates a state in which a person remains still between two actions in vector space. On the other hand, a stay indicates a state in which a person moves between two actions in vector space. In the first embodiment, (2) above is performed by the calculation unit 25, but it may also be performed by the classification unit 26.

When the consecutive action proximity is calculated in (1) above, in (2) above, the calculation unit 25 compares the calculated consecutive action proximity with a preset threshold and determines whether the two actions are a stay or a movement. For example, if the calculated consecutive action proximity is less than the threshold, the two actions are determined to be a stay, and if the calculated consecutive action proximity is equal to or greater than the threshold, the two actions are determined to be a movement.

When the continuous action time is calculated in (1) above, in (2) above, the calculation unit 25 compares the calculated time with a preset threshold and determines whether the two actions are a stay or a move. For example, if the calculated continuous action time is less than the threshold, the two actions are determined to be a stay, and if the calculated continuous action time is equal to or greater than the threshold, the two actions are determined to be a move.

When calculating the continuous action proximity and continuous action time in (1) above, in (2) above, the calculation unit 25 first derives the time derivative of the continuous action proximity from the calculated continuous action proximity and continuous action time. If the continuous action proximity is distance, the time derivative of the continuous action proximity is velocity, and if the continuous action proximity is angle, the time derivative of the continuous action proximity is angular velocity. The calculation unit 25 then compares the derived time derivative of the continuous action proximity with a preset threshold to determine whether the two actions are a stay or a move. For example, if the derived time derivative of the continuous action proximity is less than the threshold, the two actions are determined to be a stay, and if the derived time derivative of the continuous action proximity is equal to or greater than the threshold, the two actions are determined to be a move.

Furthermore, when calculating the number of stays in the behavioral history, the calculation unit 25 uses, for example, the number of actions in the behavioral history as the number of stays. In other words, if a person's actions directly indicate a stay, such as when purchasing a product, the number of actions of this person becomes the number of stays.

As described above, the number of stays calculated by the calculation unit 25 may be the number of stays between two temporally consecutive actions, or the number of actions taken by a person. In other words, the number of stays is the number of stays by a person in the action history.

The classification unit 26 classifies the behavior history into behavior transition patterns based on one or both of the start/end behavior proximity and the number of stays calculated by the calculation unit 25. A behavior transition pattern is a classification of a series of behavioral histories from when a person starts an action to when they finish the action into a predetermined pattern.

When the calculation unit 25 calculates both the start/end behavior proximity and the number of stays, the classification unit 26 classifies the behavior history into behavior transition patterns by shape based on the start/end behavior proximity and the number of stays. Specifically, the classification unit 26 regards the locations where the behavior started, the location where the behavior ended, and the locations where the behavior occurred as "points," and derives the shape by connecting these points with "lines." In other words, the shape is formed by mapping the behavior history onto a geographic space or vector space and connecting the mapped behavior histories with lines.

The visualization unit 27 visualizes the behavioral transition patterns classified by the classification unit 26.

The communication unit 28 mediates communication with the network 30 and performs data communication with the database device 10.

In the information processing device 20 of this embodiment, the program is stored in the storage unit 22, but it may also be stored in a computer-readable storage medium such as a computer-readable hard disk drive (HDD), solid-state drive (SSD), flexible disk (FD), compact disk (CD), magnetic optical disk (MO), or various types of memory. The program can also be stored in the storage medium by downloading it via a communication network such as the Internet.

<Information processing method>
Next, a description will be given of an information processing method performed using the information processing system 1 configured as above. In the following, the information processing method of the present invention will be described using an embodiment including four specific examples.

(First embodiment)
An information processing method according to a first embodiment will be described. Fig. 2 is a flow diagram showing main steps of the information processing method according to the first embodiment. In the first embodiment, a case will be described in which the behavior history data to be processed includes a usage history when a transportation IC card is used at a ticket gate in a station.

[S11: Storage process]
First, when a transportation IC card is used at a ticket gate at a station, the usage history is transmitted to the database device 10 and stored in the memory unit 12. The behavior history data stored in the memory unit 12 is input to the information processing device 20 via the network 30 and stored in the memory unit 22.

Figure 3 is a table showing an example of behavioral history data. The behavioral history data includes, for example, user ID, date and time, station, and entry and exit, which are arranged in chronological order. The memory unit 22 also stores station master data shown in Figure 4. The station master data includes geographic information about the station, i.e., latitude and longitude.

[S12: Extraction step]
Next, the extraction unit 23 extracts behavioral history from the behavioral history data stored in the storage unit 12 in step S11. In step S12, behavioral history is extracted for each identifier, as shown in FIG. 5 . In the first embodiment, the identifier is a combination of a user ID and a date. For example, a user with a user ID of 0004 performed an action on April 2 and April 3, 2020, and the identifiers are "0004_20200402" and "0004_20200403," respectively. Furthermore, the behavioral history extracted in step S12 is a temporally continuous behavioral history from the start to the end of the action. For example, for the identifier "0001_20200401," the start of the action is 10:54 and the end of the action is 14:33.

[S13: Mapping step]
Next, the mapping unit 24 maps the behavior history extracted in step S12 into geographic space. In step S13, as shown in Fig. 6, the stations in the behavior history in Fig. 5 are linked with the geographic information (latitude and longitude) of the stations in the station master data in Fig. 4. In this case, the behavior history is linked with the geographic information, and therefore the behavior history is mapped into geographic space.

Note that steps S12 and S13 may be performed in reverse order. That is, the behavioral history may be linked to geographical information, and the behavioral history may be mapped to geographic space, after which an identifier may be assigned.

[S14: Calculation step]
Next, the calculation unit 25 calculates the temporal proximity between the first and last actions in the action history using the action history mapped in step S13. In step S14, first, as shown in FIG. 7, a vector is generated for each identifier, with elements each consisting of two consecutive actions. Hereinafter, the earlier action in the two consecutive actions is referred to as the "previous action" and the later action is referred to as the "later action." The elements of the previous action and the later action include time, station, entry/exit, and geographic information of the station. Next, as the consecutive action proximity, which is the proximity between the two actions, the distance between the stations of the previous and subsequent actions is calculated for each identifier using the geographic information (latitude and longitude) of the station of the previous action and the geographic information of the station of the later action. Note that, although the first embodiment calculates the straight-line distance between the stations of the previous and subsequent actions, the distance along a geographical route may also be calculated. Next, the sum of the above distances is calculated for each identifier. Then, as shown in FIG. 8, for each identifier, the distance between the station of the first action (first station) and the station of the last action (last station) is calculated as start-end action closeness, which is the closeness between the first action and the last action.

[S15: Classification process]
Next, the classification unit 26 classifies the behavior history into behavior transition patterns based on the start-end behavior proximity calculated in step S14. In step S15, as shown in Figure 9, it is determined whether the first station and the last station are the same based on the start-end behavior proximity.

If the first and last stations are different stations, the behavioral transition pattern for that identifier is determined to be a "one-way pattern." As shown in Figure 10(a), a one-way pattern is a pattern in which the user travels from the first station to the last station without returning to the first station.

On the other hand, if the first and last stations are the same station, the behavioral transition pattern for that identifier is determined to be a "round trip pattern" or a "round trip pattern." As shown in Figure 10(b), a round trip pattern is a pattern in which you travel from the first station (departure) to the last station (destination) and then return to the first station. As shown in Figure 10(c), a round trip pattern is a pattern in which you travel from the first station to the last station via multiple stations (destinations), and more specifically, enter and exit multiple stations.

According to the first embodiment described above, behavioral history is classified into behavioral transition patterns in step S15, making it possible to appropriately understand the actual usage of each station and the flow of people. For example, when passengers travel, by distinguishing between use of transportation that returns to the departure point and use of transportation for the purpose of travel (not returning to the departure point), it is possible to hypothesize the purpose of passenger travel and tally the number of passengers. Furthermore, because people's behavioral history can be appropriately analyzed in this way, the analysis results can be used for a variety of purposes, such as product and service marketing and recommendations, event planning, and transportation planning.

Here, conventionally, OD (Origin: point of departure, Destination: destination) data has been used to analyze the flow of people. However, analysis using data from leaving a station to entering, in addition to data from entering to exiting the station (the OD data) as in the first embodiment, has not been done before. And according to the first embodiment, by classifying behavioral history into behavioral transition patterns in this way, it becomes possible to analyze people's behavioral history more appropriately than before.

In addition, as described in Patent Document 1, for example, it has been common to analyze usage trends within and around train stations based on the usage history of transportation IC cards. However, in the first embodiment, by analyzing behavioral transition patterns using data from the time a person leaves a station until the time they enter, it is possible to understand stays and movements in an area wider than the area around the station, including the area between stations. From this perspective, the first embodiment makes it possible to analyze people's behavioral history more appropriately than in the past.

In addition, in the first embodiment, behavioral history is classified into behavioral transition patterns for each identifier. By aggregating and analyzing behavioral history by identifier in this way, it is possible to prevent excessive compression of the large amount of information recorded on a transportation IC card, which makes it easier to aggregate behavioral history but prevents detailed information from being lost. It is also possible to prevent excessive expansion of large amounts of information, which makes it possible to grasp detailed information but prevents the inability to aggregate it quantitatively. In other words, the significance of classifying behavioral history into behavioral transition patterns in the first embodiment lies in the fact that aggregating and analyzing behavioral history by identifier makes it possible to strike a balance between information compression and expansion.

Furthermore, the identifier in the first embodiment is a combination of the user ID and date, and the behavioral history is classified into behavioral transition patterns for each day. By dividing time into meaningful sections in this way, a person's behavioral history can be analyzed appropriately. Note that in the first embodiment, the behavioral history is classified by day, but the time unit is not limited to this. For example, an identifier may be assigned for each trip, and the behavioral history may be classified into behavioral transition patterns.

Furthermore, in the first embodiment, behavioral history data including actions taken intentionally by an individual is analyzed, so it is meaningful to classify the behavioral history into behavioral transition patterns as human behavior. For example, analyzing automatically acquired data, such as log data acquired using GPS, and classifying behavioral transition patterns will not allow for an appropriate analysis of human behavior.

Second Embodiment
An information processing method according to the second embodiment will be described. Fig. 11 is a flow diagram showing the main steps of the information processing method according to the second embodiment. In the second embodiment, as in the first embodiment, a case will be described in which the behavior history data to be processed includes the usage history when a transportation IC card is used at a ticket gate in a station. Furthermore, in the second embodiment, the same behavior history data as in the first embodiment is used.

[S21: storage step, S22: extraction step, S23: mapping step]
Steps S21 to S23 in the second embodiment are the same as steps S11 to S13 in the first embodiment. That is, first, in step S21, the behavior history data shown in FIG. 3 and the station master data shown in FIG. 4 are stored in the storage unit 22 of the information processing device 20. Next, in step S22, behavior history is extracted for each identifier from the behavior history data as shown in FIG. 5. Next, in step S23, as shown in FIG. 6, the stations in the behavior history in FIG. 5 are linked with the geographic information of the stations in the station master data in FIG. 4, and the behavior history is mapped to geographic space.

[S24: Calculation step]
Next, the calculation unit 25 calculates the number of stays in the behavior history using the behavior history mapped in step S23. In step S24, the following three steps, step S24-1 to step S24-3, are performed to calculate the number of stays.

In step S24-1, first, as in step S24 in the first embodiment, a vector is generated for each identifier, with elements each consisting of two consecutive actions, as shown in FIG. 12. Next, the distance between the stations of the previous and following actions is calculated for each identifier as consecutive action closeness, which is the proximity between the two actions, using geographic information for the station of the previous action and the station of the following action. This calculation of consecutive action closeness is the same as in the first embodiment. In the second embodiment, the time between the previous and following actions is further used to calculate the consecutive action time, which is the time between the two actions.

Next, in step S24-2, for each identifier, the distance between two actions is determined to be a stay or a move based on the continuous action proximity and continuous action duration. Here, the threshold for continuous action proximity (distance between stations) for determining whether to stay or move is set to, for example, 3000 m, and the threshold for continuous action duration is set to 30 minutes. If the continuous action proximity is less than 3000 m and the continuous action duration is less than 30 minutes, it is determined to be a "stay," and if the continuous action proximity is 3000 m or more or the continuous action duration is 30 minutes or more, it is determined to be a "move." Note that in step S24-1, the distance between stations may be divided by the time to derive the speed, and the derived speed may be compared with a predetermined threshold to determine whether the distance between two actions is a stay or a move.

Next, in step S24-3, the number of stays is counted for each identifier and calculated as the number of stays.

In step S24 of the second embodiment, both the continuous action proximity and the continuous action time are used to determine whether the two actions are a stay or a move, but either one may be used. For example, when continuous action proximity is used, the threshold for continuous action proximity is set to 3000 m, and if the continuous action proximity is less than 3000 m, it is determined to be a "stay," and if it is 3000 m or more, it is determined to be a "move." Also, when continuous action time is used, if the continuous action time is less than 30 minutes, it is determined to be a "stay," and if it is 30 minutes or more, it is determined to be a "move."

In addition, although the determination of stay and movement in step S24 in the second embodiment is performed by the calculation unit 25, it may also be performed by the classification unit 26.

[S25: Classification process]
Next, the classification unit 26 classifies the behavior history into behavior transition patterns for each identifier based on the number of stays calculated in step S24. In step S25, as shown in Fig. 13, if the number of stays is zero, it is classified into a "no stay" behavior transition pattern. If the number of stays is one, it is classified into a "single place stay" behavior transition pattern. If the number of stays is two or more, it is classified into a "multiple place stay" behavior transition pattern.

[S26: Visualization process]
Next, the visualization unit 27 visualizes the behavioral transition patterns classified in step S25. In step S26, the number of passengers for each length of stay in the table shown in Fig. 13 is tallied for each identifier. Then, the tallied result of the number of passengers for each length of stay is visualized in a graph as shown in Fig. 14.

The second embodiment described above also provides the same effects as the first embodiment. That is, in step S25, behavioral history is classified into behavioral transition patterns, making it possible to appropriately grasp the actual usage of each station and the flow of people. Specifically, by aggregating the number of destinations where passengers stay, it is possible to grasp the purpose of passenger travel and how they stop at multiple stations. Furthermore, in step S26, behavioral transition patterns are visualized, making it even easier to grasp the actual usage of each station and the flow of people.

(Third embodiment)
An information processing method according to the third embodiment will be described. Fig. 15 is a flow diagram showing the main steps of the information processing method according to the third embodiment. In the third embodiment, as in the first and second embodiments, a case will be described in which the behavior history data to be processed is data including usage history when a transportation IC card is used at a ticket gate at a station. Furthermore, in the third embodiment, the same behavior history data as in the first and second embodiments is used.

[S31: Storage step, S32: Extraction step, S33: Mapping step]
Steps S31 to S33 in the third embodiment are the same as steps S11 to S13 in the first embodiment. That is, first, in step S31, the behavior history data shown in FIG. 3 and the station master data shown in FIG. 4 are stored in the storage unit 22 of the information processing device 20. Next, in step S32, behavior history is extracted for each identifier from the behavior history data as shown in FIG. 5. Next, in step S33, as shown in FIG. 6, the stations in the behavior history in FIG. 5 are linked with the geographic information of the stations in the station master data in FIG. 4, and the behavior history is mapped to geographic space.

[S34: Derivation step]
Next, the calculation unit 25 derives the stay station with the longest elapsed time for each identifier using the behavior history mapped in step S33. In step S34, first, as shown in FIG. 16, the relationship between two consecutive behaviors is determined as a stay or a move for each identifier. This determination of stay or move is similar to that in the second embodiment shown in FIG. 12, and the relationship between two behaviors is determined as a stay or a move based on the proximity of the consecutive behaviors and the consecutive behavior time. Next, when the relationship between two behaviors is a stay, it is determined whether the consecutive behavior time (elapsed time) is the longest. Then, the stay with the longest elapsed time is extracted, and the corresponding stay station is derived. In the example of FIG. 16, the station for which the stay with the longest elapsed time is "true" is the stay station.

In the third embodiment, the calculation unit 25 derives the station where the maximum elapsed time was spent in step S34, but this may also be done by the classification unit 26.

[S35: Visualization process]
Next, the visualization unit 27 visualizes the number of major passengers staying at each station based on the station with the longest elapsed time calculated in step S34. In step S35, the number of passengers staying at each station is calculated for each identifier, assuming that the major stations during the day are the destinations. The results of the calculation of the number of passengers staying at each station are then visualized in a graph, as shown in FIG. 17 .

According to the third embodiment described above, in step S34, the main destinations of passengers' journeys can be identified by deriving the station where the longest elapsed time has passed. As a result, the actual usage of each station and the flow of people can be properly understood.

(Fourth embodiment)
An information processing method according to the fourth embodiment will be described. Fig. 18 is a flow diagram showing main steps of the information processing method according to the fourth embodiment. In the fourth embodiment, a case will be described in which the behavior history data to be processed includes a product purchase history on an EC site. Note that the same applies to the case in which the behavior history data is a product purchase history at a shopping center rather than a product purchase history on an EC site.

[S41: Storage process]
First, when a product is purchased on an EC site, the purchase history is transmitted to the database device 10 and stored in the storage unit 12. The behavior history data stored in the storage unit 12 is input to the information processing device 20 via the network 30 and stored in the storage unit 22.

Figure 19 is a table showing an example of behavior history data. The behavior history data includes, for example, a user ID, purchase date and time, the store of purchase, and the purchased item, which are arranged in chronological order. The storage unit 22 also stores store master data shown in Figure 20 and item master data shown in Figure 21. Store topics in the store master data are created by extracting topics from the behavior history data using a topic model. Similarly, item topics in the item master data are created by extracting topics from the behavior history data using a topic model.

[S42: Extraction step]
Next, the extraction unit 23 extracts behavioral history from the behavioral history data stored in the storage unit 12 in step S41. In step S42, first, behavioral history is extracted for each user ID as shown in FIG. 22 , and then behavioral history is extracted for each identifier. The identifier in the fourth embodiment is a combination of the user ID and the date. For example, user ID 0001 performed purchasing behavior on December 1 and December 23, 2020, and the identifiers are assigned "0001" and "0002," respectively. Furthermore, the behavioral history extracted in step S42 is a temporally continuous behavioral history from the start to the end of the behavior.

[S43: Mapping step]
Next, the behavior history extracted in step S42 is vectorized and mapped to a vector space in the mapping unit 24. In step S43, the behavior history data is linked to store and item master data as shown in Fig. 23 to generate a vector whose elements are the store and item where the purchase was made.

[S44: Calculation step]
Next, the calculation unit 25 calculates the start-end behavior closeness, which is the temporal closeness between the first behavior and the last behavior in the behavior history, and the number of stays in the behavior history, using the behavior history vectorized in step S43. In step S44, the start-end behavior closeness is calculated in step S44-1, and the number of stays is calculated in step S44-2.

In step S44-1, first, the first and last actions in a series of action histories are extracted for each identifier, as shown in FIG. 24. Next, as shown in FIG. 25, for each identifier, the cosine similarity between the vectors of the previous and next actions is calculated as the start-end action closeness, which is the closeness between the first and last actions, using a vector consisting of the store topic and item topic of the previous action and a vector consisting of the store topic and item topic of the next action. Note that, although cosine similarity was calculated in the fourth embodiment, Euclidean distance or Manhattan distance may also be calculated.

In step S44-2, the number of stays is calculated as the number of actions (number of vectors) in the action history for each identifier, as shown in FIG. 26. Note that steps S44-1 and S44-2 can be performed in any order.

[S45: Classification process]
Next, the classification unit 26 classifies the behavior history into behavior transition patterns by shape based on the start-end behavior proximity and the number of stays calculated in step S44. In step S45, the behavior history is classified into five shapes according to the rules shown in FIG. 27. That is, the number of vectors, which represent the number of stays, is treated as points, and the shape is derived by connecting the points with lines. Therefore, if the number of vectors is one, the shape is a "point," if the number of vectors is two, the shape is a "line," and if the number of vectors is three or more, the shape is a "polygon." In this case, the threshold value of the cosine similarity for determining the size of the shape is set to, for example, 0.707. Then, according to these rules, the behavior history is classified into behavior transition patterns by shape for each identifier, as shown in FIG. 28.

[S46: Visualization process]
Next, the behavior transition patterns classified in step S45 are visualized in the visualization unit 27. In step S27, a shape representing the behavior transition pattern is visualized as shown in FIG.

The fourth embodiment described above can also achieve the same effects as the first and second embodiments described above. That is, because the behavioral history is classified into behavioral transition patterns in step S45, it is possible to properly understand the actual usage of each station and the flow of people. As a result, the analysis results can be used for a variety of purposes, such as marketing and recommendations for products and services.

In addition, in the fourth embodiment, behavioral histories are classified into behavioral transition patterns based on their shape. This shape has two main elements. The first element is the proximity of the start and end actions, for example, whether one end of a polygon is closed or not. This makes it possible to determine whether it is a one-way pattern or a round-trip/circumnavigation pattern. The second element is the number of stays, for example, the points on the polygon. This makes it possible to determine whether it is a round-trip pattern or a circular pattern.

Furthermore, by classifying behavioral history by shape in this way, it is possible to understand the correlation between behavioral variation and number. For example, if the goal is to increase behavioral variation, products and services can be recommended so that the number of points on the polygon is large. Furthermore, understanding the relationship between the points on the polygon and the shape can be used to recommend products and services, and to design store layouts. For example, if product A and product B go well together, product B can be recommended to purchasers of product A. Alternatively, if store C and store D go well together, it may be possible to locate store C and store D close to each other.

The above describes preferred embodiments of the present invention with reference to the accompanying drawings, but the present invention is not limited to such examples. It is clear that those skilled in the art will be able to conceive of various modifications and alterations within the scope of the ideas set forth in the claims, and it is understood that these naturally fall within the technical scope of the present invention.

This invention is useful for analyzing a person's behavioral history.

REFERENCE SIGNS LIST 1 Information processing system 10 Database device 11 Calculation unit 12 Storage unit 13 Communication unit 20 Information processing device 21 Calculation unit 22 Storage unit 23 Extraction unit 24 Mapping unit 25 Calculation unit 26 Classification unit 27 Visualization unit 28 Communication unit 30 Network

Claims (20)

The information processing device
an extraction step of extracting, from data including human behavior, a time-series behavior history from the start to the end of the behavior for each identifier that identifies an individual;
a mapping step of vectorizing the behavioral history and mapping it into a vector space ;
a calculation step of calculating , using the mapped behavioral history, one or both of the temporal proximity between the first behavior and the last behavior in the behavioral history and the number of stays in the behavioral history;
and a classification step of classifying the behavior history into behavior transition patterns based on one or both of the proximity and the number of stays. The information processing device
an extraction step of extracting, from data including human behavior, a time-series behavior history from the start to the end of the behavior for each identifier that identifies an individual;
a mapping step of mapping the behavioral history to a geographical space ;
a calculation step of calculating , using the mapped behavioral history, one or both of the temporal proximity between the first behavior and the last behavior in the behavioral history and the number of stays in the behavioral history;
and a classification step of classifying the behavior history into behavior transition patterns based on one or both of the proximity and the number of stays. In the calculation step, both the proximity and the number of stays are calculated,
3. The information processing method according to claim 1, wherein in the classifying step, the behavior history is classified into behavior transition patterns according to shape based on the proximity and the number of stays. 4. The information processing method according to claim 1 , wherein in the calculation step, the physical distance between the positions where the first action and the last action were performed is calculated as the proximity. 4. The information processing method according to claim 1 , wherein in the calculation step, a similarity between the first action and the last action is calculated as the closeness. The calculation step
calculating one or both of the proximity of two temporally consecutive actions and the time between the two actions;
determining whether the two activities are a stay or a move based on one or both of the proximity of the two activities and the time between the two activities;
The information processing method according to any one of claims 1 to 5 , further comprising the step of calculating the number of stays. 6. The information processing method according to claim 1 , wherein in the calculation step, the number of stays is calculated as the number of actions in the action history. 8. The information processing method according to claim 1 , wherein the data includes identification information of a person who performed the action. 9. The information processing method according to claim 1 , wherein the data includes data acquired from a transportation IC card or a mobile terminal. A program for causing an information processing device to execute the information processing method according to any one of claims 1 to 9 . A computer-readable storage medium storing the program according to claim 10 . An information processing device,
a storage unit that stores a program for controlling the information processing device;
a calculation unit that controls the information processing device in accordance with the program;
an extraction unit that extracts, from data including human behavior, a time-sequential behavior history from the start to the end of the behavior for each identifier that identifies an individual;
a mapping unit that vectorizes the behavior history and maps it into a vector space ;
a calculation unit that calculates , using the mapped behavior history, one or both of the temporal proximity between the first behavior and the last behavior in the behavior history and the number of stays in the behavior history;
a classification unit that classifies the behavior history into behavior transition patterns based on one or both of the proximity and the number of stays. An information processing device,
a storage unit that stores a program for controlling the information processing device;
a calculation unit that controls the information processing device in accordance with the program;
an extraction unit that extracts, from data including human behavior, a time-sequential behavior history from the start to the end of the behavior for each identifier that identifies an individual;
a mapping unit that maps the behavior history into a geographical space ;
a calculation unit that calculates , using the mapped behavior history, one or both of the temporal proximity between the first behavior and the last behavior in the behavior history and the number of stays in the behavior history;
a classification unit that classifies the behavior history into behavior transition patterns based on one or both of the proximity and the number of stays. The calculation unit calculates both the proximity and the number of stays,
The information processing device according to claim 12 or 13 , wherein the classifying unit classifies the behavior history into behavior transition patterns according to shapes, based on the proximity and the number of stays. The information processing device according to claim 12 , wherein the calculation unit calculates the proximity as a physical distance between positions where the first action and the last action were performed. The information processing device according to claim 12 , wherein the calculation unit calculates a similarity between the first action and the last action as the closeness. The calculation unit
Calculating one or both of the proximity of two consecutive actions and the time between the two actions;
determining whether the two activities are a stay or a move based on one or both of the proximity of the two activities and the time between the two activities;
The information processing device according to claim 12 , wherein the number of stays is calculated. The information processing device according to claim 12 , wherein the calculation unit calculates the number of the actions in the action history as the number of stays. The information processing device according to claim 12 , wherein the data includes identification information of a person who performed the action. The information processing device according to any one of claims 12 to 19 , wherein the data includes data acquired from a transportation IC card or a mobile terminal.