Information processing device and information processing method
The information processing device uses a learning model to minimize the distance between feature quantities of the same user and maximize the distance between different users, addressing the challenge of inaccurate data matching by enhancing user data association accuracy.
Patent Information
- Authority / Receiving Office
- JP · JP
- Patent Type
- Applications
- Current Assignee / Owner
- KDDI CORP
- Filing Date
- 2024-12-10
- Publication Date
- 2026-06-22
AI Technical Summary
Existing techniques struggle to accurately match data of the same user when a large number of users are present at the same position, leading to inaccuracies in data association.
An information processing device and method that utilizes a learning model to output feature quantities from location history data, trained to minimize the distance between features of the same user and maximize the distance between different users, enabling accurate matching through a triplet loss function and clustering based on related data.
Enables high-accuracy matching of first and second data of the same user by leveraging a learning model that distinguishes between similar and different user behaviors, improving data association accuracy.
Smart Images
Figure 2026101011000001_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to an information processing apparatus and an information processing method.
Background Art
[0002] Conventionally, data of the same user included in a first data group collected by a first operator and a second data group collected by a second operator has been associated. For example, in Patent Document 1, the first data group includes first data of a plurality of users, and the second data group includes second data of a plurality of users. Based on whether or not the position of the user indicated by the position history information included in the first data approximates the position of the user indicated by the position history information included in the second data, the first data and the second data are matched.
Prior Art Documents
Patent Documents
[0003]
Patent Document 1
Summary of the Invention
Problems to be Solved by the Invention
[0004] When a large number of users are present at the same position, the technique described in Patent Document 1 has a problem that the first data and the second data of the same user cannot be accurately matched.
[0005] Therefore, the present invention has been made in view of these points, and an object thereof is to enable accurate matching of the first data and the second data of the same user.
Means for Solving the Problems
[0006] An information processing device according to a first aspect of the present invention includes: a location history data acquisition unit that acquires a first data group including a plurality of first location history data associated with first user identification information for identifying each of a plurality of first users collected by a first business operator and the location history of each of the first users; a second data group including a plurality of second location history data associated with second user identification information for identifying each of a plurality of second users collected by a second business operator and the location history of each of the second users; and a learning model that outputs a feature quantity indicating the characteristics of the location history shown by the location history data in response to the input of the location history data, wherein the positive example distance, which is the distance between the feature quantity output for the input of the first location history data corresponding to an anchor user which is a reference user and the feature quantity output for the input of the second location history data of the same user as the anchor user, is the distance between the feature quantity output for the input of the first location history data which is the distance between the feature quantity output for the input of the second location history data of the same user as the anchor user and the previous user corresponding to the anchor user The system includes a feature acquisition unit that acquires first feature quantities, which are the feature quantities of each of the multiple first users, by inputting the first location history data of each of the multiple first users into the learning model, which has been trained so that the feature quantities output for input of first location history data are smaller than the negative example distance, which is the distance between the feature quantities output for input of second location history data of a user different from the anchor user, and the learning model that acquires first feature quantities, which are the feature quantities of each of the multiple first users, by inputting the second location history data of each of the multiple second users into the learning model, and a matching unit that matches the multiple first location history data with the multiple second location history data based on the distance between the first feature quantities of each of the multiple first users and the second feature quantities of each of the multiple second users acquired by the feature acquisition unit.
[0007] The learning model may be a model that has learned a triplet loss that maximizes the distance between the positive example distance and the negative example distance as its loss function.
[0008] The location history data included in the first or second data group is associated with related data related to the location history, and the learning model may be a learning model that has been further trained so that the similarity of two features output for each input of two location history data corresponding to two related data belonging to different clusters is smaller than the similarity of two features output for each input of two location history data belonging to two related data belonging to the same cluster when the first or second data group is clustered based on the related data.
[0009] A user different from the anchor user may be a user whose associated data belongs to a different cluster than the cluster to which the associated data corresponding to the anchor user belongs.
[0010] A second aspect of the present invention relates to an information processing method, which is executed by a computer, comprising the steps of acquiring a first data group comprising a plurality of first location history data associated with first user identification information for identifying each of a plurality of first users collected by a first business operator and the location history of each of the first users, and a second data group comprising a plurality of second location history data associated with second user identification information for identifying each of a plurality of second users collected by a second business operator and the location history of each of the second users, and a learning model that outputs a feature quantity indicating the characteristics of the location history data shown by the location history data in response to the input of the location history data, wherein the positive example distance, which is the distance between the feature quantity output for the input of the first location history data corresponding to an anchor user which is a reference user and the feature quantity output for the input of the second location history data of the same user as the anchor user, is the same as the anchor The method includes the steps of: obtaining a first feature, which is the feature of each of the plurality of first users, by inputting the first location history data of each of the plurality of first users into the learning model, which has been trained so that the first feature output for input of first location history data corresponding to a user is smaller than the negative example distance, which is the distance between the feature output for input of second location history data of a user different from the anchor user; and obtaining a second feature, which is the feature of each of the plurality of second users, by inputting the second location history data of each of the plurality of second users into the learning model; and matching the plurality of first location history data with the plurality of second location history data based on the distance between the obtained first feature of each of the plurality of first users and the second feature of each of the plurality of second users. [Effects of the Invention]
[0011] According to the present invention, the first and second data of the same user can be matched with high accuracy. [Brief explanation of the drawing]
[0012] [Figure 1]This is a diagram illustrating the overview of an information processing device. [Figure 2] This diagram shows the functional configuration of an information processing device. [Figure 3] This figure shows examples of the first and second data sets. [Figure 4] This is a flowchart showing the processing flow in an information processing device. [Modes for carrying out the invention]
[0013] [Overview of Information Processing Device 1] Figure 1 is a diagram illustrating the overview of the information processing device 1. The information processing device 1 is a computer that matches first location history data, which shows the location history of multiple first users, with second location history data, which shows the location history of multiple second users.
[0014] Information processing device 1 is operated by, for example, an aggregation service provider that aggregates data and provides the aggregated data, and is connected to external devices such as the first device 2 and the second device 3 via a communication network (not shown), such as the Internet or a mobile phone line.
[0015] The first device 2 is, for example, a computer operated by the first business operator. The first device 2 manages a first data set that includes multiple first location history data sets, which associate first user IDs (Identification), which are first user identification information used to identify each of the multiple first users collected by the first business operator, with each first user's location history. The second device 3 is, for example, a computer operated by the second business operator. The second device 3 manages a second data set that includes multiple second location history data sets, which associate second user IDs, which are second user identification information used to identify each of the multiple second users collected by the second business operator, with each second user's location history.
[0016] Among the plurality of second users and the plurality of first users, there are also the same users. However, the first user identification information corresponding to the same user is different from the second user identification information, and it is assumed that the location history data of the same user cannot be matched by performing matching between the first user identification information and the second user identification information.
[0017] The information processing device 1 acquires a first data group from the first device 2 and a second data group from the second device 3. The information processing device 1 can utilize a learning model that outputs a feature amount indicating the feature of the location history indicated by the location history data for the input of the location history data in order to perform matching between the first location history data and the second location history data.
[0018] The learning model is a learning model that is learned so that a positive example distance, which is the distance between the feature amount output for the input of the first location history data corresponding to an anchor user, which is a reference user, and the feature amount output for the input of the second location history data of the same user as the anchor user, is smaller than a negative example distance, which is the distance between the feature amount output for the input of the first location history data corresponding to the anchor user and the feature amount output for the input of the second location history data of a user different from the anchor user.
[0019] The information processing apparatus 1 obtains a first feature amount, which is a feature amount of each of a plurality of first users, by inputting the first position history data of each of the plurality of first users into the learning model, and obtains a second feature amount, which is a feature amount of each of a plurality of second users, by inputting the second position history data of each of the plurality of second users into the learning model. Then, the information processing apparatus 1 performs matching between the plurality of first position history data and the plurality of second position history data based on the distances between the first feature amounts of each of the plurality of first users and the second feature amounts of each of the plurality of second users. Thus, by performing matching based on the feature amounts output from the learning model that is trained so that the distance between the first feature amount and the second feature amount of the same user is small and the distance between the first feature amount and the second feature amount of different users is large, the information processing apparatus 1 can accurately match the first data and the second data of the same user.
[0020] [Functional Configuration of Information Processing Apparatus 1] Next, the functional configuration of the information processing apparatus 1 will be described. FIG. 2 is a diagram showing the functional configuration of the information processing apparatus 1.
[0021] As shown in FIG. 2, the information processing apparatus 1 includes a communication unit 11, a storage unit 12, and a control unit 13. The communication unit 11 is a communication interface for transmitting and receiving data to and from the first device 2, the second device 3, etc. via a communication network.
[0022] The storage unit 12 is a storage medium for storing various data, and includes a ROM (Read Only Memory), a RAM (Random Access Memory), a hard disk, an SSD (Solid State Drive), a flash memory, and the like. The storage unit 12 stores a program executed by the control unit 13. The storage unit 12 stores a program that causes the control unit 13 to function as a position history data acquisition unit 131, a generation unit 132, a feature amount acquisition unit 133, a matching unit 134, and an output unit 135.
[0023] The control unit 13 is, for example, a CPU (Central Processing Unit). By executing a program stored in the memory unit 12, the control unit 13 functions as a location history data acquisition unit 131, a generation unit 132, a feature quantity acquisition unit 133, a matching unit 134, and an output unit 135.
[0024] The location history data acquisition unit 131 acquires a first data group containing multiple first location history data sets, each associated with a first user ID for identifying each of the multiple first users collected by the first business operator and the location history of each first user, and a second data group containing multiple second location history data sets, each associated with a second user ID for identifying each of the multiple second users collected by the second business operator and the location history of each second user.
[0025] Figure 3 shows an example of the first and second data groups. Figure 3(A) shows the first data group, and Figure 3(B) shows the second data group. As shown in Figure 3(A), the first location history data, which associates the first user ID with the location history, is associated with related data related to the location history, specifically the number of times the first user has visited location A and the number of times the first user has visited location B, as identified based on the location history. Similarly, as shown in Figure 3(B), the second location history data, which associates the second user ID with the location history, is associated with related data related to the location history, specifically data indicating whether the first user, identified based on the location history, visited a supermarket and whether the second user visited a park.
[0026] Here, location history is information that associates, for example, a date and time with latitude and longitude information indicating the user's location corresponding to that date and time, showing the location for multiple dates and times. Related data, on the other hand, is data with fewer dimensions than location history.
[0027] Furthermore, in this embodiment, as shown in Figures 3(A) and (B), the first user ID and the second user ID are data with different systems, and are different from the first and second user IDs of the same user. As a result, it is not possible to match the location history data of the same user by associating the first and second user IDs.
[0028] The generation unit 132 generates a learning model that outputs feature quantities representing the characteristics of the location history data in response to the input location history data. For example, the generation unit 132 generates a learning model by performing a first learning process and a second learning process.
[0029] The first learning process involves reducing the distance between the features of the same user in the first location history data and the second location history data, while increasing the distance between the features of different users. For example, the generation unit 132 receives a specification from a user terminal (not shown) used by a user performing the integration of the first and second data sets, for the first location history data corresponding to the anchor user, which is the reference user, from among the first location history data of multiple users included in the first data set. The generation unit 132 also receives a specification from the user for the second location history data corresponding to the anchor user, from among the second location history data of multiple users included in the second data set, as well as a specification for second location history data corresponding to a user different from the anchor user. The first and second location history data corresponding to the anchor user are called positive example data, and the combination of the first location history data corresponding to the anchor user and the second location history data corresponding to other users is called negative example data.
[0030] The generation unit 132 learns the model such that the positive example distance, which is the distance between the feature output for input of first location history data corresponding to an anchor user and the feature output for input of second location history data of the same user as the anchor user, is smaller than the negative example distance, which is the distance between the feature output for input of first location history data corresponding to the anchor user and the feature output for input of second location history data of a different user than the anchor user. Specifically, the generation unit 132 sets the loss function L1 to a triplet loss that maximizes the distance between the positive example distance and the negative example distance, and performs the first learning so that the triplet loss output by the loss function L1 is minimized.
[0031] Here, F is the number of features that the learning model outputs for input of the first location history data corresponding to the anchor user, and F is the number of features that the learning model outputs for input of the second location history data of the same user as the anchor user. + F is the number of features that the learning model outputs for input of second location history data of a user different from the anchor user. - Let m be the distance margin. The generation unit 132 then trains the learning model so that the loss function L1 shown in equation (1) below is minimized when multiple positive and negative example data are input to the learning model to obtain features. Here, the d function is a distance function, for example, the Euclidean distance. The max function is a function that returns the largest number among the multiple numbers included in the max function.
number
[0032] The second type of learning involves reducing the distance between the features of multiple users who behave similarly within either the first or second dataset, while increasing the distance between the features of multiple users who behave differently. An example of performing the second type of learning using the first dataset is described below.
[0033] For example, the generation unit 132 clusters the first data group based on the related data included in the first data group. Then, the generation unit 132 performs a second learning process such that the similarity of the two features output for each input of two location history data corresponding to two related data belonging to different clusters becomes smaller than the similarity of the two features output for each input of two location history data corresponding to two related data belonging to the same cluster when the first data group was clustered based on the related data. Here, the greater the similarity of the two features, the more similar the two features are considered to be.
[0034] Specifically, P(i) is the set of indices of users in the same cluster as the i-th user, A(i) is the set of indices of all users except the i-th user, and z i , z p , z a These are the i-th, p-th, and a-th user features, respectively, and τ is a temperature parameter used to adjust the control loss output by the loss function. The generation unit 132 then trains the learning model so that the control loss output by the loss function L2, shown in equation (2) below, is minimized.
[0035]
number
[0036] Furthermore, when performing the first learning, it is preferable that users other than the anchor user are users whose related data belongs to a different cluster than the cluster to which the related data corresponding to the anchor user belongs. In response to this, the generation unit 132 receives a specification from the user terminal of the user performing the integration of the first data group and the second data group for the second location history data corresponding to the anchor user from among the second location history data of multiple users included in the second data group, and also receives a specification for second location history data corresponding to users who belong to a different cluster than the cluster to which the anchor user belongs, when the related data associated with the second location history data is clustered as users other than the anchor user.
[0037] In this way, the information processing device 1 can learn during the first learning phase using negative example data based on the features of other users who behave differently from the anchor user, and learn in such a way that the difference between the features of this anchor user and those of other users becomes apparent. This improves learning efficiency compared to learning using negative example data based on the features of users similar to the anchor user.
[0038] Furthermore, if the location history information is from a space where multiple users with similar hobbies and interests gather, such as an event venue, it may be difficult to discern differences in the behavior of multiple users. To address this, when performing the first learning process, users different from the anchor user may be defined as users whose related data belongs to the same cluster as the related data corresponding to the anchor user. In this way, the information processing device 1 can learn to differentiate features even among multiple users who behave similarly.
[0039] The feature acquisition unit 133 acquires first features, which are the features of each of the multiple first users, by inputting the first location history data of each of the multiple first users acquired by the location history data acquisition unit 131 into the learning model generated by the generation unit 132. Furthermore, the feature acquisition unit 133 acquires second features, which are the features of each of the multiple second users, by inputting the second location history data of each of the multiple second users acquired by the location history data acquisition unit 131 into the learning model generated by the generation unit 132.
[0040] The matching unit 134 performs matching between multiple first location history data and multiple second location history data based on the distance between the first feature vectors of each of the multiple first users acquired by the feature acquisition unit 133 and the second feature vectors of each of the multiple second users.
[0041] For example, the matching unit 134 identifies the second feature that is closest to each of the multiple first features acquired by the feature acquisition unit 133. Then, the matching unit 134 determines that the first user of the first location history data corresponding to the first feature and the second user of the second location history data corresponding to the second feature are the same, provided that the distance between the first feature and the second feature identified for that first feature is below a predetermined threshold. As a result of matching, it is possible that multiple second users may correspond to one first user. In this case, the matching unit 134 may resolve this problem by matching the features in such a way that the transport distance is minimized.
[0042] The output unit 135 outputs integrated data by associating the first location history data and related data of the first user, which is determined to be the same user based on the results of the matching unit 134, with the second location history data and related data of the second user. For example, the output unit 135 outputs the integrated data to the user terminal used by the user who is integrating the first data group and the second data group.
[0043] [Operation Flow] Next, we will explain the processing flow related to the information processing device 1. Figure 4 is a flowchart showing the processing flow in the information processing device 1. First, the location history data acquisition unit 131 acquires a first data group from the first device 2 and a second data group from the second device 3 (S1).
[0044] Next, the generation unit 132 receives the first position history data of the anchor user and the second position history data of the anchor user as positive example data from the user terminal of the user using the information processing device 1, and also receives the first position history data of the anchor user and the second position history data of a user different from the anchor user as negative example data (S2).
[0045] Next, the generation unit 132 performs a first learning process based on the positive example data and the negative example data, and then performs a second learning process to generate a learning model (S3). In this flowchart, the processes S2 and S3 are performed, but this is not the only way. If a learning model has already been generated, the information processing device 1 does not need to perform the processes related to S2 and S3.
[0046] Next, the feature acquisition unit 133 inputs the first location history data of each of the multiple first users into the generated learning model to acquire first features, which are the features of each of the multiple first users, and inputs the second location history data of each of the multiple second users into the learning model to acquire second features, which are the features of each of the multiple second users (S4).
[0047] Next, the matching unit 134 performs matching between multiple first location history data and multiple second location history data based on the distance between the first feature of each of the multiple first users and the second feature of each of the multiple second users (S5).
[0048] Next, the output unit 135 outputs integrated data based on the matching results from the matching unit 134 (S6). Specifically, the output unit 135 generates integrated data by associating the first location history data of the first user and the second location history data of the second user, which have been determined to be identical, and outputs the integrated data to the user terminal.
[0049] [Effects of Information Processing Device 1] As described above, in the information processing system S according to this embodiment, the information processing device 1 acquires first features, which are the features of each of the multiple first users, by inputting the first location history data of each of the multiple first users, and acquires second features, which are the features of each of the multiple second users, by inputting the second location history data of each of the multiple second users, by inputting the first location history data of each of the multiple first users, and acquires second features, which are the features of each of the multiple second users, by inputting the second location history data of each of the multiple second users, by inputting the first features of each of the multiple first users, and the second features, which are the features of each of the multiple second users, by inputting the second location history data of each of the multiple second users, by inputting the first features of each of the multiple first users, and the second features, which are the features of each of the multiple second users, by inputting the second location history data of each of the multiple second users, and the information processing device 1 then matches the multiple first location history data and the multiple second location history data based on the distance between the acquired first features of each of the multiple first users and the second features of each of the multiple second users. In this way, the information processing device 1 can accurately match the first data and second data of the same user.
[0050] Furthermore, this invention makes it possible to contribute to Goal 9 of the United Nations-led Sustainable Development Goals (SDGs), "Build resilient infrastructure, promote inclusive and sustainable industrialization and foster innovation." In addition, all or part of the device can be configured by functionally or physically distributing and integrating it in any unit. Moreover, new embodiments resulting from any combination of multiple embodiments are also included in the embodiments of this invention. The effects of the new embodiments resulting from the combinations are combined with the effects of the original embodiments. [Explanation of Symbols]
[0051] 1. Information Processing Device 2 1st device 3 Second device 11 Communications Department 12 Storage section 13 Control Unit 131 Location history data acquisition unit 132 Generation part 133 Feature acquisition unit 134 Matching Section 135 Output section
Claims
1. A location history data acquisition unit acquires a first data group which includes a plurality of first location history data associated with first user identification information for identifying each of a plurality of first users collected by a first business operator and the location history of each of the first users, and a second data group which includes a plurality of second location history data associated with second user identification information for identifying each of a plurality of second users collected by a second business operator and the location history of each of the second users, A learning model that outputs feature quantities representing the characteristics of location history data in response to input location history data, wherein the positive example distance, which is the distance between the feature quantity output for input of first location history data corresponding to an anchor user (a reference user) and the feature quantity output for input of second location history data of the same user as the anchor user, is trained to be smaller than the negative example distance, which is the distance between the feature quantity output for input of first location history data corresponding to the anchor user and the feature quantity output for input of second location history data of a different user than the anchor user, and a feature quantity acquisition unit that acquires first feature quantities, which are the characteristics of each of the multiple first users, by inputting the first location history data of each of the multiple first users into the learning model, and acquires second feature quantities, which are the characteristics of each of the multiple second users, by inputting the second location history data of each of the multiple second users into the learning model, A matching unit performs matching between the plurality of first location history data and the plurality of second location history data based on the distance between the first feature quantity of each of the plurality of first users acquired by the feature quantity acquisition unit and the second feature quantity of each of the plurality of second users. An information processing device having
2. The aforementioned learning model is a model that has learned a triplet loss that maximizes the distance between the positive example distance and the negative example distance as its loss function. The information processing apparatus according to claim 1.
3. The location history data included in the first data group or the second data group is associated with related data related to the location history, The learning model is a learning model that has been further trained so that the similarity of two feature quantities output for each input of two location history data corresponding to two related data belonging to different clusters is smaller than the similarity of two feature quantities output for each input of two location history data corresponding to two related data belonging to different clusters, when the first data group or the second data group is clustered based on the related data. The information processing apparatus according to claim 1.
4. A user different from the aforementioned anchor user is a user whose associated data belongs to a different cluster than the cluster to which the associated data corresponding to the aforementioned anchor user belongs. The information processing apparatus according to claim 3.
5. A computer executes Steps to obtain: a first data group including multiple first location history data associated with first user identification information for identifying each of multiple first users collected by a first business operator and the location history of each of the first users; and a second data group including multiple second location history data associated with second user identification information for identifying each of multiple second users collected by a second business operator and the location history of each of the second users; A learning model that outputs feature quantities representing the characteristics of location history shown by location history data in response to input location history data, wherein the positive example distance, which is the distance between the feature quantity output for input of first location history data corresponding to an anchor user (a reference user) and the feature quantity output for input of second location history data of the same user as the anchor user, is trained such that the positive example distance is smaller than the negative example distance, which is the distance between the feature quantity output for input of first location history data corresponding to the anchor user and the feature quantity output for input of second location history data of a different user than the anchor user, and the learning model is trained to input the first location history data of each of the plurality of first users to obtain first feature quantities, which are the features of each of the plurality of first users, and the learning model is trained to input the second location history data of each of the plurality of second users to obtain second feature quantities, which are the features of each of the plurality of second users, A step of matching the plurality of first location history data with the plurality of second location history data based on the distance between the first feature quantity of each of the plurality of first users and the second feature quantity of each of the plurality of second users, An information processing method having