Position information management device and position information management method

Abstract

In the present invention, an acquisition unit acquires first position information indicating the position of a first user at a first time and third position information indicating the position of the first user at a third time later than the first time. A determination unit determines second position information indicating the position of the first user at a second time between the first time and the third time, by processing the first position information and the third position information while using an estimation model which has learned, on the basis of movement history information of a group of first users, a staying location at which each user in the group of first users stayed between two discretionary locations. A presentation unit presents information based on the second position information.

Description

Location Information Management Device and Location Information Management Method 【0001】 The present invention relates to a location information management device and a location information management method. 【0002】 Conventionally, the utilization of trip chain (Trip Chain) data, which is time-series data of an individual's location information, has been promoted. For example, the traffic behavior estimation system described in Patent Document 1 below estimates traffic behavior selection in consideration of a plurality of continuous action plans. The traffic behavior estimation system integrally uses the result of estimating the traffic behavior of an individual based on a plurality of movement histories included in the trip chain of the individual and the result of estimating the traffic behavior of the individual based on each movement history of the individual to estimate the traffic behavior of the individual. 【0003】 International Publication No. 2016 / 067460 【0004】 As an example of a method for generating trip chain data, it may be possible to obtain the location information of a user over time by receiving a signal transmitted from a communication terminal possessed by the user at a base station of a mobile phone network. In such a case, in a situation where the communication status between the base station and the communication terminal is poor, there may be a gap in the trip chain data, and the trip chain data may become sparse. When the trip chain data is sparse, it may not be possible to represent the original movement of the user, and the accuracy of processing using the trip chain data may decrease. 【0005】 An object of the present invention is to make sparse trip chain data non-sparse. 【0006】A location information management device according to one aspect of the present invention includes: an acquisition unit that acquires first location information indicating the location of a first user at a first time and third location information indicating the location of the first user at a third time after the first time; a determination unit that determines second location information indicating the location of the first user at a second time between the first time and the third time by processing the first location information and the third location information using an estimation model that has learned the locations where each user in the first user group stayed between any two points based on the movement history information of the first user group; and a presentation unit that presents information based on the second location information. 【0007】 A location information management method according to one aspect of the present invention acquires first location information indicating the location of a first user at a first time, and third location information indicating the location of the first user at a third time after the first time. The method processes the first location information and the third location information using an estimation model that has learned the locations where each user in the first user group stayed between any two points based on the movement history information of the first user group, thereby determining second location information indicating the location of the first user at a second time between the first time and the third time, and presenting information based on the second location information. 【0008】 According to one aspect of the present invention, sparse trip chain data can be made non-sparse. 【0009】 This is a diagram showing the configuration of the location information management system 1 according to an embodiment. This is a block diagram showing the configuration of the user terminal 20-1. This is a block diagram showing the configuration of the location information management device 10. This is a schematic diagram showing an example of the contents of the location information database DB1. This is a diagram showing an example of the contents of the trip chain data TD1. This is a diagram showing an example of the contents of the trip chain data TD2. This is a diagram showing an example of the contents of the trip chain data TD2. This is a schematic diagram showing the processing by the estimation model LM. This is a flowchart showing the operation of the processing device 105 of the location information management device 10. This is a diagram showing an example of the contents of the trip chain data TD2-4 according to the first modified example. 【0010】[Embodiment] [System Configuration] Figure 1 shows the configuration of a location information management system 1 according to an embodiment. The location information management system 1 includes a location information management device 10 and a plurality of user terminals 20 (20-1 to 20-n (where n is an integer of 2 or more)). Each of the user terminals 20-1 to 20-n is used by a different user U (U-1 to U-n). The location information management device 10 and the user terminals 20-1 to 20-n are connected via a network N. 【0011】 User terminals 20-1 to 20-n are, for example, information processing terminals such as smartphones or tablet devices. In this embodiment, user terminals 20-1 to 20-n are to be carried and used by users U-1 to U-n. Therefore, the location information of user terminals 20-1 to 20-n can be considered as the location information of users U-1 to U-n. 【0012】 The location information management device 10 manages the location information of user terminals 20-1 to 20-n. The location information management device 10 periodically acquires the location information of each of the multiple user terminals 20-1 to 20-n and generates trip chain data TD for users U-1 to U-n (user terminals 20-1 to 20-n). The location information management device 10 also vectorizes the trip chain data TD and generates trip chain vector TV. As will be described in detail later, the trip chain vector TV is stored in a database and used for behavioral analysis or behavioral prediction of users U-1 to U-n. 【0013】 For example, to determine the location of user terminal 20-1, there are several methods, such as the following: Method 1: Periodically acquire location information generated by the GPS (Global Positioning System) device 204 (see Figure 2) of user terminal 20-1. Method 2: Consider the location information of the base station with which user terminal 20-1 communicates among the base stations of the mobile phone network as the location information of user terminal 20-1. Method 3: Consider the location information of the access point with which user terminal 20-1 communicates among the Wi-Fi® access points installed in the city as the location information of user terminal 20-1. In this embodiment, the location information management device 10 acquires the location information of user terminals 20-1 to 20-n using Method 1. 【0014】 [User Terminals 20-1 to 20-n] Figure 2 is a block diagram showing the configuration of user terminal 20-1. Since the configurations of user terminals 20-1 to 20-n are substantially the same, Figure 2 will use user terminal 20-1 as an example for explanation. User terminals 20-1 to 20-n include a display device 201, an input device 202, a communication device 203, a GPS device 204, a storage device 205, a processing device 206, and a bus 220 that connects these devices to each other. 【0015】 The display device 201 is a display device that displays information to the outside (for example, various display panels such as liquid crystal display panels or organic EL display panels). The input device 202 is an input device that accepts input from the outside (for example, a keyboard, mouse, microphone, switch, button, or sensor). The display device 201 and the input device 202 may be configured as an integrated unit (for example, a touch panel). The communication device 203 has an interface that can be connected to the network N and communicates with other devices connected to the network N using wireless or wired communication. 【0016】 The GPS device 204 receives radio waves from multiple satellites and calculates location information indicating the position of the user terminal 20-1 from the received radio waves. The location information can be in any format as long as it can identify the position of the user terminal 20-1. In this embodiment, the GPS device 204 calculates latitude and longitude as location information. 【0017】 The frequency (period) at which the GPS device 204 generates location information is arbitrary, but generating location information at a high frequency will cause the battery level of the user terminal 20-1 to decrease more easily, and the amount of communication in the entire location information management system 1 will also increase. Therefore, the GPS device 204 generates location information at a frequency set by user U-1, or at a frequency that does not hinder tracking user U-1's movement history. 【0018】The storage device 205 is a recording medium that can be read by the processing device 206. The storage device 205 includes, for example, non-volatile memory and volatile memory. Non-volatile memory includes, for example, ROM (Read Only Memory), EPROM (Erasable Programmable Read Only Memory), and EEPROM (Electrically Erasable Programmable Read Only Memory). Volatile memory includes, for example, RAM (Random Access Memory). The storage device 205 stores program PG2. Program PG2 is a program for operating the user terminal 20. 【0019】 The processing unit 206 includes one or more CPUs (Central Processing Units). One or more CPUs are examples of one or more processors. Each of the processors and CPUs is an example of a computer. 【0020】 The processing unit 206 reads the program PG2 from the storage device 205. By executing the program PG2, the processing unit 206 functions as a location information transmission unit 211. The location information transmission unit 211 may be composed of circuits such as a DSP (Digital Signal Processor), an ASIC (Application Specific Integrated Circuit), a PLD (Programmable Logic Device), and an FPGA (Field Programmable Gate Array). 【0021】 The location information transmission unit 211 transmits the location information generated by the GPS device 204 to the location information management device 10 via the communication device 203. The location information transmission unit 211 transmits terminal information to the location information management device 10, which is an association between identification information that identifies the user terminal 20-1, the location information of the user terminal 20-1, and time information indicating the time when the location information was obtained. The time when the location information was obtained is approximately the same as the time when the location information was generated. 【0022】The timing of transmitting terminal information is arbitrary. The location information transmission unit 211 may, for example, transmit terminal information including location information each time location information is generated by the GPS device 204. Alternatively, the location information transmission unit 211 may, for example, store terminal information since the previous terminal information transmission time in the storage device 205 until a predetermined terminal information transmission time arrives, and then transmit the terminal information stored in the storage device 205 when the terminal information transmission time arrives. Furthermore, the location information included in the terminal information may be all of the location information generated by the GPS device 204, or it may be location information generated by the GPS device 204 that has been thinned out at predetermined intervals. In the latter case, for example, while the GPS device 204 generates location information every second, the location information included in the terminal information may be thinned out every 10 seconds. 【0023】 [Location Information Management Device 10] Figure 3 is a block diagram showing the configuration of the location information management device 10. The location information management device 10 includes a display device 101, an input device 102, a communication device 103, a storage device 104, a processing device 105, and a bus 120 that connects these devices to each other. 【0024】 The display device 101 is a display device that displays information to the outside (for example, various display panels such as liquid crystal display panels or organic EL display panels). The input device 102 is an input device that accepts input from the outside (for example, a keyboard, mouse, microphone, switch, button, or sensor). The display device 101 and the input device 102 may be configured as an integrated unit (for example, a touch panel). The communication device 103 has an interface that can be connected to the network N and communicates with other devices connected to the network N using wireless or wired communication. 【0025】The storage device 104 is a recording medium that can be read by the processing device 105. The storage device 104 includes, for example, non-volatile memory and volatile memory. The non-volatile memory is, for example, ROM, EPROM, and EEPROM. The volatile memory is, for example, RAM. The storage device 104 stores a program PG1, a location information database DB1, a vector database DB2, a user database DB3, and an estimation model LM. 【0026】 Program PG1 is a program for operating the location information management device 10. The location information database DB1 stores location information (terminal information as described above) that is continuously acquired from user terminals 20-1 to 20-n. The vector database DB2 stores trip chain vector TV, which is a vectorized version of the trip chain data TD generated based on the location information. The user database DB3 stores information about users U-1 to U-n. Information about users U-1 to U-n includes, for example, attribute information of users U-1 to U-n. Attribute information of users U-1 to U-n includes, for example, information such as the occupation, age, gender, and place of residence of users U-1 to U-n. 【0027】 The estimation model LM is a machine learning model trained to estimate the location of users U-1 to U-n at a given time based on their previous and subsequent locations, when location information for user terminals 20-1 to 20-n cannot be obtained and the location of users U-1 to U-n at a given time in the past cannot be determined. Details of the estimation model LM will be described later. 【0028】 The processing unit 105 includes one or more CPUs. One or more CPUs are examples of one or more processors. Each of the processors and CPUs is an example of a computer. By executing the program PG1, the processing unit 105 functions as a learning control unit 110, an acquisition unit 111, a generation unit 112, a determination unit 113, an output unit 114, an extraction unit 115, and a presentation unit 116. At least some of these functions may be configured by circuits such as a DSP, ASIC, PLD, and FPGA. 【0029】The acquisition unit 111 acquires location information (the terminal information described above) from each user terminal 20-1 to 20-n. As described above, the acquisition of location information by the acquisition unit 111 is performed continuously at predetermined intervals. The acquisition unit 111 records the location information acquired from user terminals 20-1 to 20-n in the location information database DB1. 【0030】 Figure 4 is a schematic diagram showing an example of the contents of the location information database DB1. The location information database DB1 includes user ID 401 and location information 402. User ID 401 is identification information that can uniquely identify user terminals 20-1 to 20-n (users U-1 to U-n). In the example in Figure 4, the user ID of user U-1 is denoted as "001", the user ID of user U-2 as "002", and the user ID of user U-3 as "003". Location information 402 is the location information (latitude and longitude) for each user U-1 to U-n. In the example in Figure 4, location information 402 for a two-hour period from 12:00 to 14:00 on December 1, 2024 is shown at 15-minute intervals. 【0031】 For example, user U-1 (user U with user ID 001) is located at the location indicated by location information (x11, y11) from 12:00 to 12:30, at the location indicated by location information (x12, y12) from 12:45 to 13:15, and returns to the location indicated by location information (x11, y11) after 13:30. Also, for example, user U-3 (user U with user ID 003) is continuously located at the location indicated by location information (x31, y31) from 12:00 to 14:00. 【0032】 Furthermore, for example, user U-2 (user U with user ID 002) is located at the location indicated by location information (x21, y21) from 12:00 to 12:30, and at the location indicated by location information (x22, y22) from 14:00 onwards, but location information between 12:45 and 13:45 could not be obtained, making it impossible to determine the location where the user was staying. Reasons why location information could not be obtained include, for example, poor communication between the user terminal 20-2 used by user U-2 and the base station. 【0033】The generation unit 112 generates trip chain data TD based on location information. Trip chain data TD is data that shows the transition of locations where users U-1 to U-n stayed during a predetermined period (for example, 3 hours from 12:00 to 15:00, 24 hours from 0:00 to 24:00, or 3 days from December 1st to December 3rd, etc.). 【0034】 For example, when generating trip chain data TD for user U-1 over a 24-hour period from 0:00 to 24:00 on December 1st, the generation unit 112 obtains user U-1's location information from the location information database DB1 for the period from 0:00 to 24:00 on December 1st. The generation unit 112 determines that user U-1 was staying at the location indicated by the location information value (latitude and longitude) if the value does not change for a predetermined determination period or longer. 【0035】 For example, let's explain how to generate trip chain data TD for the two-hour period from 12:00 to 14:00 using the location information of user U-1 (user ID 001) in Figure 4. If the above determination period is 30 minutes, it is determined that user U-1 was staying at the location indicated by the value if the location information acquired at 15-minute intervals is the same value for two consecutive times. In this case, user U-1 was at the location indicated by location information (x11, y11) at 12:00, at the location indicated by location information (x12, y12) at 12:45, and returned to the location indicated by location information (x11, y11) at 13:30. The location indicated by location information (x11, y11) is user U-1's office, and the location indicated by location information (x12, y12) is the cafeteria. In this case, it can be estimated that user U-1 worked in the office until after 12:30, ate lunch in the cafeteria during their lunch break, and returned to the office by 13:30. 【0036】Figure 5 shows an example of the contents of trip chain data TD1. The trip chain data TD1 shown in Figure 5 is generated using the location information of user U-1 in Figure 4. Trip chain data TD1 includes user ID 501, time 502, and location 503. Time 502 is the time when the location information of the location indicated by location 503 was first acquired. Location 503 is information indicating the location where user U-1 stayed. In Figure 5, location 503 is indicated by facility names such as "office" and "cafeteria," but in actual trip chain data TD1, location 503 is indicated by a mesh ID. A mesh ID is a numerical value (or string) assigned to each mesh in which a map is divided into a mesh of a predetermined distance on each side, so as to be uniquely identifiable. In this embodiment, the mesh ID is indicated as "C + 3 digits." 【0037】 The determination unit 113 desparsens the trip chain data TD if it is sparse. If location information cannot be obtained due to factors such as poor communication status of user terminals 20-1 to 20-n, the location of users U-1 to U-n cannot be determined, and the trip chain data TD may become sparse. The determination unit 113 desparsens the trip chain data TD by having the estimation model LM estimate the location of users U-1 to U-n during the period when location information could not be obtained. 【0038】 Figures 6A to 6C show an example of the contents of trip chain data TD2. The trip chain data TD2-1 shown in Figure 6A is generated using the location information of user U-4 (user with user ID 004), which is not shown in Figure 4. Like trip chain data TD1, trip chain data TD2-1 also includes user ID 501, time 502, and location 503. 【0039】Referring to the trip chain data TD2-1, user U-4 was at "home" at 8:20, at the "supermarket" (labeled "super" in the diagram) at 8:30, and at the "cafe" at 15:00. More than seven hours passed between user U-4's stay at the "supermarket" and their stay at the "cafe," but since location information during this period could not be obtained, the specific locations of stay could not be determined. In other words, the trip chain data TD2-1 is sparse data. 【0040】 Therefore, the determination unit 113 specifies an arbitrary time (hereinafter referred to as the "specified time") within the period during which location information could not be obtained, such as the trip chain data TD2-2 shown in Figure 6B, and causes the estimation model LM to estimate the location (place of stay) of user U-4 at that time. The specified time may be specified, for example, by the administrator of the location information management device 10 (location information administrator), or it may be automatically determined by the processing unit 105. In the example in Figure 6B, 12:00, which is approximately the middle time during the period during which location information could not be obtained, is set as the specified time. In addition, the special token "<MASK>" is entered at the place of stay 503 corresponding to the specified time. 【0041】 In other words, the determination unit 113 generates quasi-sparse trip chain data TD2-2 by adding supplementary data DH1 to sparse trip chain data TD2-1. Quasi-sparse trip chain data TD is trip chain data TD that includes special tokens. The position in which the supplementary data DH1 is added is determined so that the time is in ascending order. The user ID 501 of the supplementary data DH1 is entered as the user ID of user U (user ID 004 in the case of Figure 6B). The time 502 is entered as the specified time (the time at which the location of user U's stay is to be predicted). The location 503 is entered as the special token "<MASK>". 【0042】When the quasi-sparse trip chain data TD2-2 is input to the estimation model LM, the estimation model LM outputs an estimated value for <MASK>. The determination unit 113 determines the estimated value output by the estimation model LM as the location where user U-4 is staying at the specified time, and generates non-sparse trip chain data TD2-3. Specifically, for example, suppose the location where user U-4 is staying at the specified time is predicted to be "Office (C256)". In this case, the determination unit 113 replaces the "<MASK>" part in the trip chain data TD2-2 with "Office (C256)", as shown in the trip chain data TD2-3 in Figure 6C. Through the above processing, the sparse trip chain data TD2-1 (see Figure 6A) is converted into non-sparse trip chain data TD2-3 (see Figure 6C). 【0043】 In summary, the acquisition unit 111 acquires first location information indicating the location of user U-4 at 8:30 and third location information indicating the location of user U-4 at 15:00, which is after 8:30. The determination unit 113 uses the estimation model LM to process the first and third location information and determines second location information indicating the location of user U-4 at 12:00, which is between 8:30 and 15:00. User U-4 is an example of the first user. 8:30 is an example of the first time. 15:00 is an example of the third time. 12:00 is an example of the second time. The first location information is, for example, the data in the second row of the trip chain data TD2-1. The third location information is, for example, the data in the third row of the trip chain data TD2-1. The second location information is, for example, the data in the third row of the trip chain data TD2-3. 【0044】 Furthermore, in the example described above, when estimating the location of user U-4, a time is specified, and the location of user U-4 at that specified time (specified time) is estimated. In other words, the acquisition unit 111 acquires the first location information and the third location information, along with the specification of 12:00 as the specified time. The determination unit 113 determines the location information of the point where user U-4 is estimated to be located at 12:00 as the second location information. 【0045】 [Estimation Model LM] Next, the details of the estimation model LM will be described. The estimation model LM is a model obtained by improving CTLE (Context and Time Aware Location Embedding) for missing value filling processing. A general CTLE is a model that estimates future stay locations using past data. In general, CTLE does not handle future data in principle. When future data is input, a general CTLE treats it as having low reliability and suppresses it from flowing into the transformer encoder TE (see Fig. 7). 【0046】 In contrast, the estimation model LM according to the present embodiment estimates the data to be complemented using, in addition to past data, future data as viewed from the specified time (the time to be estimated). By using future data, the complementation accuracy is improved. Also, there is no limit to the number of past data and future data used for estimation. There is also no temporal constraint on the data used for estimation. For example, it is not limited to relatively recent data such as the day before or two days before the specified time, and data from, for example, one month ago may be used. 【0047】 Fig. 7 is a schematic diagram showing the processing by the estimation model LM. The estimation model LM includes an encoding layer (Encoding Layer) EL, a transformer encoder (Transformer Encoder) TE, and downstream models (Downstream Models) DL. First, time-series data D1 to D4 are input to the encoding layer EL. The time-series data D1 to D4 are, for example, the trip chain data TD2-2 shown in Fig. 6B. More specifically, let the first row of the trip chain data TD2-2 be data D1, the second row be data D2, the third row be data D3', and the fourth row be data D4. The data D3' includes the special token "<MASK>". 【0048】The encoding layer EL converts the data D1 to D4 into a format processable by the transformer encoder TE. In FIG. 7, the converted data is data Z'1 to Z'4. The data Z'1 to Z'4 is input to the transformer encoder TE. The transformer encoder TE outputs data Z1 to Z4 corresponding to the data Z'1 to Z'4. The data Z1 to Z4 is input to the downstream model DM, and finally, data D3, which is the estimated value of the special token "<MASK>", is output. 【0049】 Here, the details of the transformer encoder TE will be described. A transformer is one of the deep learning models for improving the performance of AI (Artificial Intelligence). The transformer includes an encoder and a decoder. The encoder is a processing unit that converts data into vectors, and the decoder is a processing unit that converts vectors into the original data format. The transformer encoder TE corresponds to the encoder part of the transformer. 【0050】 The transformer encoder TE can perform high-speed and accurate natural language processing by learning using an attention layer. The attention layer is a processing unit that determines which words are important information in a sentence. For example, when translating and outputting an English sentence "This is a pen." into a Japanese sentence "これはペンです。", it learns which of the words "This", "is", "a", "pen", "." are important. In this sentence example, the importance of "pen" is the highest. 【0051】 In the present embodiment, the transformer encoder TE generates the features (vectors) of the data to be estimated based on the degree of relevance (weight) between the input data. As described above, the input data is data D1, data D2, data D3' and data D4. Among these, data D1 and D2 are past data with respect to the specified time, data D3' is the data at the specified time (the data to be estimated), and data D4 is future data with respect to the specified time. 【0052】 For example, considering the correlation between data D1 and data D3', data D1's location is home, but since most people start their activities from home, this does not provide a clue to estimating data D3's location. In other words, the correlation between data D1 and data D3' is estimated to be low. Considering the correlation between data D2 and data D3', data D2's location is a supermarket, but people who go to the supermarket in the morning (8:30) tend to go straight to work. Therefore, the correlation between data D2 and data D3' is estimated to be high. Considering the correlation between data D4 and data D3', people who go to a cafe at 15:00 may be on a work break. Therefore, the correlation between data D4 and data D3' is estimated to be moderate. 【0053】 Next, we will explain the training of the estimation model LM. The estimation model LM is pre-trained using MLM (Masked Language Modeling). MLM is a learning method used in deep learning models that perform natural language processing to understand the context of a text. The procedure for pre-training using MLM is as follows: Step 1: Prepare multiple texts to be used for pre-training. Step 2: Select one of the multiple texts and divide the selected text into tokens. Step 3: Replace a certain percentage (e.g., 20%) of the tokens in the selected text with a special token (e.g., <MASK>). Step 4: Input the text with some of the words replaced with special tokens into the deep learning model and have it estimate the words before they were replaced with special tokens. Step 5: If the estimation result is incorrect, adjust the various parameters of the deep learning model to improve the estimation accuracy. Repeat steps 2 to 5 below. 【0054】When a sentence containing <MASK> is input to a deep learning model trained in this way, the deep learning model outputs an estimated value for <MASK>. Specifically, for example, if the sentence "My favorite sport is <MASK>, and I would like to watch it in America someday" is input to the deep learning model, since this sentence contains tokens such as "sports," "America," and "watch," the deep learning model predicts that <MASK> is "baseball" and outputs "baseball" as the predicted value. 【0055】 In this embodiment, the training of the estimation model LM is performed by the training control unit 110. The procedure for training the estimation model LM is as follows: Step 1: Prepare multiple training trip chain data TDs to be used for pre-training. The training trip chain data TDs are sparse trip chain data TDs (or non-sparsed trip chain data TDs). The training trip chain data TDs may be data generated based on the location information of users U-1 to U-n, or data generated based on the location information of a group of users different from users U-1 to U-n. Step 2: Select one of the multiple trip chain data TDs and replace some of the stay locations in the selected trip chain data TD with special tokens (e.g., <MASK>). Step 3: Input the trip chain data TD with some of the locations replaced with special tokens into the estimation model LM and have it estimate the stay locations before replacement with special tokens. Step 4: If the estimation result is incorrect, adjust various parameters of the estimation model LM to improve the estimation accuracy. Repeat steps 2 through 4 below. 【0056】 Thus, the estimation model LM learns the locations where each user U in the first user group stayed between any two points, based on the tripchain data TD of the first user group. The first user group consists of any user U from U-1 to U-n. The tripchain data TD is an example of travel history information. The locations where a user stayed between any two points are locations that have been replaced with special tokens. 【0057】For example, let's describe the case where the trip chain data TD1 of user U-1 shown in Figure 5 is used as training data. In this case, the first user group includes user U-1, and the trip chain data TD of the first user group includes the trip chain data TD1 of user U-1. The trip chain data TD1 of user U-1 includes a fourth location information (office (C211)) indicating the location of user U-1 at 12:00, a fifth location information (cafeteria (C223)) indicating the location of user U-1 at 12:45 after 12:00, and a sixth location information (office (C211)) indicating the location of user U-1 at 13:30 after 12:45. The learning control unit 110 uses the trip chain data TD1 with the fifth location information masked as training data and trains the estimation model LM by having the estimation model LM estimate the fifth location information. User U-1 is an example of a second user. 12:00 is an example of the fourth time zone. 12:45 is an example of the fifth time zone. 13:30 is an example of the sixth time zone. 【0058】 The output unit 114 outputs a trip chain vector TV based on the trip chain data TD. The trip chain vector TV is represented by multidimensional numerical values ​​such as (0.53, -1.58, 2.20, 0.01, 0.89, ..., -1.13). The output unit 114 converts the trip chain data TD into a trip chain vector TV using, for example, Doc2vec. Doc2vec is a processing model for converting sequential data such as sentences into vector representations. If the Doc2vec model is properly trained, the cosine similarity of vectors generated from two sentences with similar meanings will be high. Doc2vec is often used in natural language processing, but it can also be applied to trip chain data TD, which is a type of sequential data. 【0059】Trip chain data TD is data that is close to natural language and difficult to perform numerical calculations on. By converting trip chain data TD into trip chain vector TV, the output unit 114 can perform quantitative processing (e.g., similarity evaluation) on user U's travel history information. In addition, trip chain data TD includes privacy-related data such as user ID and places of stay. By converting trip chain data TD into trip chain vector TV, this privacy-related data can be masked (converted into a form in which the content of the data cannot be directly identified), thereby improving data security. The trip chain vector TV output from the output unit 114 is stored in the vector database DB2. The output unit 114 may also be referred to as the vector generation unit. 【0060】 For example, the output unit 114 outputs a trip chain vector TV1 showing the movement history of user U-4 based on the trip chain data TD2-3 shown in Figure 6C, that is, based on the first position information, the second position information, and the third position information. The trip chain vector TV1 is an example of the first movement vector. 【0061】 When a trip chain vector TV is specified, the extraction unit 115 extracts other trip chain vector TVs similar to the specified trip chain vector TV from the vector database DB2. Hereinafter, the specified trip chain vector TV will be referred to as the "specified vector". The extraction unit 115 may also be described as a "vector search unit" that searches for similar vectors similar to the specified vector. 【0062】 The extraction unit 115 extracts trip chain vector TVs with high cosine similarity to a specified vector from among the trip chain vector TVs stored in the vector database DB2. Cosine similarity is a measure of how similar two vectors are. The extraction unit 115 extracts, for example, trip chain vector TVs with a similarity to the specified vector of a predetermined value or higher, or a predetermined number of trip chain vector TVs in descending order of similarity to the specified vector, from the vector database DB2. 【0063】 Furthermore, the extraction unit 115 may extract attribute information of user U whose movement history is identified by the similarity vector from the user database DB3. User U whose movement history is identified by the specified vector and user U whose movement history is identified by the similarity vector can be said to have similar behavioral characteristics. By analyzing the attribute information of these users U, it is possible to efficiently perform tasks such as selecting target audiences for advertisements or selecting the content of region-specific campaigns. 【0064】 For example, the extraction unit 115 extracts at least one similar vector from the vector database DB2, which shows the movement history of users U-1 to U-n, that is similar to the trip chain vector TV1 output based on the trip chain data TD2-3 shown in Figure 6C. Users U-1 to U-n are examples of the second user group. The vector database DB2 is an example of a movement vector database. 【0065】 Furthermore, the extraction unit 115 extracts attribute information of at least one user U whose movement history is identified by at least one similar vector from the user database DB3, which stores attribute information for each of the users U-1 to U-n. Each of the users U-1 to U-n is an example of a user U included in the second user group. The user database DB3 is an example of an attribute information database. 【0066】The presentation unit 116 presents information based on the non-sparsed trip chain data TD. In this embodiment, the presentation unit 116 presents information based on the second location information. The information based on the second location information may be, for example, the non-sparsed trip chain data TD of user U-4, which is completed by the completion of the second location information. In this case, the presentation unit 116 presents the trip chain data TD2-3 of user U-4, which includes the second location information. Presentation may be, for example, display on the display device 101, audio output from a speaker (not shown), or printed output using a printer (not shown). When the second location information is determined by the determination unit 113, the presentation unit 116 may automatically display the trip chain data TD2-3 on the display device 101. This allows the location information manager to confirm the trip chain data TD2-3 of user U-4, in which the location at the specified time has been completed. 【0067】 Furthermore, the presentation may also refer to output from other information processing terminals, such as user terminals 20-1 to 20-n. In this case, the presentation unit 116 transmits information based on the second location information and control information for outputting information based on the second location information to the other information processing terminal. 【0068】Furthermore, the presentation unit 116 may present information regarding the trip chain vector TV corresponding to the trip chain data TD that has been made non-sparse by the second location information. The information regarding the trip chain vector TV may be, for example, the number of similar vectors when the trip chain vector TV is designated as a specified vector. This allows, for example, a location information administrator to determine whether the movement history identified by the designated vector is a typical movement behavior. Alternatively, the information regarding the trip chain vector TV may be, for example, attribute information of a user U whose movement history is identified by a similar vector. In this case, the presentation unit 116 presents attribute information of at least one user U whose movement history is identified by a similar vector. This allows, for example, a location information administrator to know the attribute information of a user U who takes a movement behavior similar to the movement history identified by the designated vector. 【0069】 [Flowchart] Figure 8 is a flowchart showing the operation of the processing unit 105 of the location information management device 10. In the flowchart of Figure 8, the presentation unit 116 presents attribute information of user U whose movement history is identified by similar vectors. The processing unit 105 functions as an acquisition unit 111 and acquires location information from user terminals 20-1 to 20-n and stores the location information in the location information database DB1 (step S100). Until the generation of trip chain data TD (indicated as "data" in the figure) is instructed (step S102: NO), the processing unit 105 returns to step S100 and continues to acquire location information. 【0070】When the generation of trip chain data TD is instructed (step S102: YES), the processing unit 105 functions as a generation unit 112 and generates trip chain data TD based on location information stored in the location information database DB1 (step S103). If there is sparse data in the trip chain data TD (step S104: YES), the processing unit 105 functions as a determination unit 113 and makes the trip chain data TD non-sparse by predicting the location of stay at the time when location information could not be obtained (specified time) and supplementing the trip chain data TD (step S105). If there is no sparse data (step S104: NO), the processing unit 105 moves the processing to step S106. 【0071】 The processing unit 105 functions as an output unit 114 and outputs a trip chain vector TV based on non-sparse trip chain data TD, and stores the trip chain vector TV in the vector database DB2 (step S106). The processing unit 105 functions as an extraction unit 115 and extracts similar vectors similar to the specified vector 1 from the vector database DB2 (step S107). At this time, the extraction unit 115 extracts attribute information of user U corresponding to the similar vector from the user database DB3. Then, the processing unit 105 functions as a presentation unit 116 and presents information about the similar vector (step S108). As described above, the presentation of information about the similar vector may be, for example, displayed on the display device 101, or transmitted to user terminals 20-1 to 20-n or other information processing terminals. After that, the processing unit 105 returns to step S100. 【0072】[Summary of Embodiments] As described above, when the location information management device 10 according to the embodiment acquires first location information indicating the location of user U-4 at 8:30 and third location information indicating the location of user U-4 at 15:00, it uses the estimation model LM to determine second location information indicating the location of user U-4 at 12:00. This makes it possible to make sparse TD data non-sparse, and to accurately grasp the actions of user U-4. 【0073】 Conventional trip chain estimation estimates the location of a person at a given time based on past trip chain data TD relative to the time of estimation. In contrast, the location information management device 10 uses future trip chain data TD relative to the time of estimation, in addition to the past trip chain data TD, to perform the estimation, thereby enabling a more accurate estimation of the location of a person at a given time. 【0074】 Furthermore, the location information management device 10 outputs a trip chain vector TV1 showing the movement history of user U-4 based on non-sparsed trip chain data TD2-3 (including first, second, and third location information). This generates a trip chain vector TV1 that accurately represents the original actions of user U-4. In addition, by representing the movement history of each user U-1 to U-n with trip chain vector TV instead of location information, computational processing such as similarity calculation can be performed. Moreover, since the trip chain vector TV does not directly indicate user ID and place of stay, the privacy of users U-1 to U-n can be protected. 【0075】 Furthermore, the location information management device 10 extracts similar vectors to the trip chain vector TV1 from the vector database DB2. The trip chain vector TV1 is vectorized data of non-sparsed trip chain data TD2-3. Therefore, compared to extracting similar vectors using trip chain vector TV based on sparse trip chain data TD, the extraction of similar vectors can be performed with higher accuracy. 【0076】 Furthermore, the location information management device 10 presents attribute information of user U whose movement history is identified by similar vectors. Therefore, for example, a location information manager can obtain attribute information of user U with a similar movement history and use the trip chain data TD for various purposes such as advertising or market research. 【0077】 Furthermore, the location information management device 10 presents the user U-4's trip chain data TD2-3, which includes the second location information. Therefore, for example, the location information manager can check the non-sparsed user U-4's trip chain data TD2-3 and perform accurate behavioral analysis of user U. 【0078】 Furthermore, the location information management device 10 uses MLM to train the estimation model LM. Therefore, it can accurately train the trip chain data TD, which is sequential data. 【0079】 [Modifications] The following are examples of modifications to the above-described embodiment. Two or more modifications can be arbitrarily selected from the following examples and combined as appropriate, provided they do not contradict each other. 【0080】 [First Modification] In the above-described embodiment, when estimating the trip chain data TD, a time is specified, and the position of user U at the specified time (specified time) is estimated. However, it is not limited to this; for example, the position of user U may be specified, and the time at which user U was located at the specified position may be estimated. 【0081】Figure 9 shows an example of the contents of trip chain data TD2-4 according to the first modification. The trip chain data TD2-4 shown in Figure 9 includes supplementary data DH2 in the third row. In the supplementary data DH2, time 502 is entered as "<MASK>" and location 503 is entered as "Office (C256)". When such quasi-sparse trip chain data TD2-4 is input to the estimation model LM, the estimation model LM outputs an estimated value of <MASK> (time 502). The determination unit 113 determines the estimated value output by the estimation model LM as the time when user U-4 stayed at the location, and generates non-sparse trip chain data TD2-3 (see Figure 6C). 【0082】 In other words, in the first modified example, the acquisition unit 111 acquires the designation of the second location information in addition to the first and third location information. The determination unit 113 determines the second time as the time at which user U-4 is estimated to have been located at the point identified by the second location information. 【0083】 According to the first modification, for example, if the location where user U-4 stayed is known but the time of stay is unknown, it is possible to estimate that time. Also, according to the first modification, for example, it is possible to estimate what time user U-4 is most likely to be at a particular location. 【0084】 [Second Modification] In the above-described embodiment, the location information management device 10 was provided with a learning control unit 110 that trained an estimation model LM. However, the estimation model LM may be trained by another information processing device. That is, the location information management device 10 may acquire the trained estimation model LM. Also, for example, the estimation model LM is not limited to being stored in the location information management device 10, but may also be stored in another information processing device that can be connected to the location information management device 10 via the network N. According to the second modification, the processing load of the location information management device 10 can be reduced. 【0085】[Third Modification] In the above-described embodiment, the location information management device 10 acquires location information from user terminals 20-1 to 20-n and generates trip chain data TD and trip chain vector TV using the location information. However, it is not limited to this, and each user terminal 20-1 to 20-n may generate trip chain data TD and trip chain vector TV from its own location information and transmit the trip chain vector TV to the location information management device 10. In this case, each of the user terminals 20-1 to 20-n may store the estimated model LM, or an information processing device that can be connected to the user terminals 20-1 to 20-n via the network N may store the estimated model LM. 【0086】 According to the third modification, the processing load on the location information management device 10 can be reduced. Furthermore, according to the third modification, since location information is not transmitted externally from user terminals 20-1 to 20-n, the privacy of users U-1 to U-n is protected. 【0087】 [Fourth Modification] The above-described embodiment explained the case in which attribute information of user U, whose movement history is identified by similar vectors, is used. However, the invention is not limited to this, and for example, the following information obtained when acquiring location information can be used to analyze user U's behavior: date, weekday / holiday, day of the week, weather, purpose of movement, means of movement, and user IDs of users who acted together. 【0088】 [Other] (1) In the embodiments described above, ROM and RAM were given as examples of storage devices 104 and 205, but storage devices 104 and 205 may be flexible disks, magneto-optical disks (e.g., compact disks, digital multipurpose disks, Blu-ray® disks), smart cards, flash memory devices (e.g., cards, sticks, key drives), CD-ROMs (Compact Disc-ROMs), registers, removable disks, hard disks, floppy® disks, magnetic strips, databases, servers, or other suitable storage media. 【0089】(2) In the embodiments described above, the information, signals, etc. may be represented using any of the various different techniques. For example, the data, instructions, commands, information, signals, bits, symbols, chips, etc. that may be referred to throughout the above description may be represented by voltage, current, electromagnetic waves, magnetic fields or magnetic particles, optical fields or photons, or any combination thereof. 【0090】 (3) In the embodiments described above, the input and output information may be stored in a specific location (e.g., memory) or managed using a management table. The input and output information may be overwritten, updated, or appended to. The output information may be deleted. The input information may be transmitted to other devices. 【0091】 (4) In the embodiments described above, the determination may be made by a value represented by one bit (0 or 1), by a Boolean value (true or false), or by a numerical comparison (for example, a comparison with a predetermined value). 【0092】 (5) The processing procedures, sequences, flowcharts, etc., exemplified in the embodiments described above may be rearranged in order, as long as there is no inconsistency. For example, in the methods described herein, various step elements are presented using an exemplary order and are not limited to the specific order presented. 【0093】 (6) Each function illustrated in Figures 2 and 3 is implemented by any combination of at least one of hardware and software. Furthermore, the method of implementing each function block is not particularly limited. That is, each function block may be implemented using one device that is physically or logically coupled, or it may be implemented using two or more physically or logically separated devices that are directly or indirectly connected (for example, using wired, wireless, etc.). A function block may also be implemented by combining the one or more devices with software. 【0094】(7) The programs illustrated in the embodiments described above should be broadly interpreted to mean instructions, instruction sets, code, code segments, program code, programs, subprograms, software modules, applications, software applications, software packages, routines, subroutines, objects, executable files, execution threads, procedures, functions, etc., whether they are called software, firmware, middleware, microcode, hardware description languages ​​or by any other name. 【0095】 Furthermore, software, instructions, information, etc., may be transmitted and received via a transmission medium. For example, if software is transmitted from a website, server, or other remote source using at least one of wired technology (such as coaxial cable, fiber optic cable, twisted pair, or digital subscriber line (DSL)) and wireless technology (such as infrared or microwave), then at least one of these wired and wireless technologies is included in the definition of a transmission medium. 【0096】 (8) In each of the above-mentioned forms, the terms “system” and “network” shall be used interchangeably. 【0097】 (9) The information, parameters, etc. described in this disclosure may be expressed using absolute values, relative values ​​from a given value, or other corresponding information. 【0098】(10) In the embodiments described above, the portable device may be a Mobile Station (MS). A Mobile Station may also be referred to by those skilled in the art as a subscriber station, mobile unit, subscriber unit, wireless unit, remote unit, mobile device, wireless device, wireless communication device, remote device, mobile subscriber station, access terminal, mobile terminal, wireless terminal, remote terminal, handset, user agent, mobile client, client, or several other appropriate terms. In this disclosure, terms such as “mobile station,” “user terminal,” “user equipment (UE),” and “terminal” may be used interchangeably. 【0099】 (11) In the embodiments described above, the terms “connected,” “coupled,” or any variation thereof means any direct or indirect connection or coupling between two or more elements, and may include the presence of one or more intermediate elements between two elements that are “connected” or “coupled” with each other. The coupling or connection between elements may be physical, logical, or a combination thereof. For example, “connection” may be read as “access.” As used in the present disclosure, two elements may be considered to be “connected” or “coupled” with each other using at least one of one or more wires, cables, and printed electrical connections, and, in some non-limiting and non-exclusive examples, electromagnetic energy having wavelengths in the radio frequency domain, microwave domain, and optical (both visible and invisible) domain. 【0100】 (12) In the embodiments described above, the phrase "based on" does not mean "based solely on" unless otherwise specified. In other words, the phrase "based on" means both "based solely on" and "based at least on". 【0101】(13) The terms “determinating” and “deciding” as used in this disclosure may encompass a wide variety of actions. “Determinating” and “deciding” may include, for example, judging, calculating, computing, processing, deriving, investigating, looking up, search, inquiry (for example, searching in a table, database or other data structure), and confirming. Furthermore, "judgment" and "decision" may include considering something as a "judgment" or "decision" based on actions such as receiving (e.g., receiving information), transmitting (e.g., sending information), input, output, and access (e.g., accessing data in memory). Additionally, "judgment" and "decision" may include considering something as a "judgment" or "decision" based on actions such as resolving, selecting, choosing, establishing, and comparing. In short, "judgment" and "decision" may include considering something as a "judgment" or "decision" based on some action. Furthermore, "judgment (decision)" may be reinterpreted as "assuming," "expecting," or "considering." 【0102】 (14) Where the terms “include,” “including,” and variations thereof are used in the embodiments described above, these terms are intended to be inclusive, as is the term “comprising.” Furthermore, the term “or” as used in this disclosure is not intended to be exclusive OR. 【0103】 (15) In the present disclosure, if articles are added by translation, for example, a, an, and the in English, the present disclosure may include the fact that the noun following these articles is plural. 【0104】(16) In this disclosure, the term “A and B are different” may mean “A and B are different from each other.” The term may also mean “A and B are each different from C.” Terms such as “separate” and “combine” may be interpreted in the same way as “different.” 【0105】 (17) Each aspect / embodiment described herein may be used individually, in combination, or switched between as needed during implementation. Furthermore, notification of certain information (e.g., notification that "it is X") is not limited to explicit notification, but may also be implicit (e.g., by not providing such notification). 【0106】 1...Location information management system, 10...Location information management device, 20 (20-1 to 20-n)...User terminal, 101...Display device, 102...Input device, 103...Communication device, 104...Storage device, 105...Processing device, 110...Learning control unit, 111...Acquisition unit, 112...Generation unit, 113...Decision unit, 114...Output unit, 115...Extraction unit, 116...Presentation unit, 201...Display device, 202...Input device, 203...Communication device, 204...GPS device, 205...Storage device, 206...Processing device, 211...Location information transmission unit, LM...Estimation model, N...Network, U (U-1 to U-n)...User.

Claims

1. A location information management device comprising: an acquisition unit that acquires first location information indicating the location of a first user at a first time point and third location information indicating the location of the first user at a third time point after the first time point; a determination unit that determines second location information indicating the location of the first user at a second time point between the first time point and the third time point by processing the first location information and the third location information using an estimation model that has learned the locations where each user in the first user group stayed between any two points based on the movement history information of the first user group; and a presentation unit that presents information based on the second location information.

2. The location information management device according to claim 1, further comprising an output unit that outputs a first movement vector indicating the movement history of the first user based on the first location information, the second location information, and the third location information.

3. The location information management device according to claim 2, further comprising an extraction unit for extracting at least one similar vector similar to the first movement vector from a movement vector database including a group of movement vectors showing the movement history of a second user group.

4. The location information management device according to claim 3, wherein the extraction unit extracts attribute information of at least one user whose movement history is identified by the at least one similar vector from an attribute information database storing attribute information of each user included in the second user group, and the presentation unit presents the attribute information of the at least one user.

5. The location information management device according to claim 1, wherein the display unit displays the movement history information of the first user, including the second location information.

6. The location information management device according to claim 1, further comprising: a first user group including a second user; the movement history information of the first user group including the movement history information of the second user; the movement history information of the second user including a fourth location information indicating the second user's position at a fourth time, a fifth location information indicating the second user's position at a fifth time after the fourth time, and a sixth location information indicating the second user's position at a sixth time after the fifth time, with the fifth location information masked, the second user's movement history information being used as training data, and a learning control unit that trains the estimation model by having the estimation model estimate the fifth location information.

7. The location information management device according to claim 1, wherein the acquisition unit acquires the first location information and the third location information, as well as the designation of the second time, and the determination unit determines the location information of the point where the first user is estimated to be located at the second time as the second location information.

8. The location information management device according to claim 1, wherein the acquisition unit acquires the designation of the second location information in addition to the first location information and the third location information, and the determination unit determines the second time to be the time at which the first user is estimated to have been located at the point identified by the second location information.

9. A location information management method comprising: acquiring first location information indicating the location of a first user at a first time point and third location information indicating the location of the first user at a third time point after the first time point; processing the first and third location information using an estimation model that has learned the locations where each user in the first user group stayed between any two points based on the movement history information of the first user group, thereby determining second location information indicating the location of the first user at a second time point between the first and third time points; and presenting information based on the second location information.

Images