Program, terminal, method, and method for generating a trained model

Abstract

This program provides an improved accuracy in determining the user's mode of transportation. [Solution] The program causes a computer to perform the following: acquire step count data representing the number of steps taken by the user holding the terminal each hour; acquire location data representing the location of the terminal at each point in time; calculate feature data representing features related to the change between locations at each point in time based on the location data; and use a trained model generated by machine learning, with training step count data representing the number of steps taken each hour and training feature data representing features related to the change between locations at each point in time as input data, and labels representing the means of transportation as training data, to estimate the user's means of transportation based on the step count data and feature data.

Description

[Technical Field]

[0001] The present invention relates to a program, a terminal, a method, and a method for generating a trained model. [Background technology]

[0002] Patent Document 1 discloses a support system that facilitates the maintenance of individual users' health and the realization of a decarbonized society by mediating an exchange agreement between users for decarbonization points, which are based on the amount of carbon dioxide emissions reduced by users refraining from using automobiles and motorcycles and instead traveling on foot or by bicycle, and benefits provided by a client. [Prior art documents] [Patent Documents]

[0003] [Patent Document 1] International Publication No. 2023 / 074552 [Overview of the Initiative] [Problems that the invention aims to solve]

[0004] The above-mentioned literature describes a method for determining the user's mode of transportation by calculating the user's movement speed and comparing it to a threshold; however, this method may not always provide sufficient accuracy in determining the mode of transportation.

[0005] The present invention has been made in view of the above problems, and its main objective is to provide a program, terminal, method, and method for generating a trained model that can improve the accuracy of determining the user's means of transportation. [Means for solving the problem]

[0006] To solve the above problems, a program according to one aspect of the present invention causes a computer to perform the following actions: acquire step count data representing the number of steps taken by a user who possesses a terminal every hour; acquire location data representing the location of the terminal at each point in time; calculate feature data representing feature quantities related to the change between the locations at each point in time based on the location data; and use a trained model generated by machine learning, with training step count data representing the number of steps taken every hour and training feature data representing feature quantities related to the change between the locations at each point in time as input data, and labels representing means of transportation as training data, to estimate the means of transportation of the user based on the step count data and the feature data.

[0007] Furthermore, a terminal according to another aspect of the present invention is a terminal held by a user, comprising: a step count acquisition unit that acquires step count data representing the user's step count per hour; a location acquisition unit that acquires location data representing the terminal's location at each point in time; a preprocessing unit that calculates feature data representing feature quantities relating to changes between locations at each point in time based on the location data; and an estimation unit that takes learning step count data representing the step count per hour and learning feature data representing feature quantities relating to changes between locations at each point in time as input data, and uses a trained model generated by machine learning, with labels representing means of transportation as training data, to estimate the user's means of transportation based on the step count data and the feature data.

[0008] Furthermore, another aspect of the present invention involves acquiring step count data representing the number of steps taken by a user holding a terminal per hour, acquiring location data representing the location of the terminal at each point in time, calculating feature data representing feature quantities related to the change between the locations at each point in time based on the location data, and using the training step count data representing the number of steps taken per hour and the training feature data representing feature quantities related to the change between the locations at each point in time as input data, and using a trained model generated by machine learning as training data, to estimate the user's means of transportation based on the step count data and the feature data.

[0009] Further, a method for generating a trained model according to another aspect of the present invention acquires step count data representing the number of steps of a moving person for each time, acquires position data representing the position of the moving person at each time point, calculates feature data representing a feature amount related to a change between the positions at each time point based on the position data, assigns a label representing the moving means of the moving person to the step count data and the feature data, and generates the trained model by machine learning using the step count data and the feature data as input data and the label representing the moving means as teacher data. [Effects of the Invention]

[0010] According to the present invention, it is possible to improve the determination accuracy of a user's moving means. [Brief Description of the Drawings]

[0011] [Figure 1] It is a diagram showing an example of a support system. [Figure 2] It is a diagram showing an example of a support device. [Figure 3] It is a diagram showing an example of a terminal. [Figure 4] It is a diagram showing an example of a terminal. [Figure 5] It is a diagram showing an example of step count data. [Figure 6] It is a diagram showing an example of position data. [Figure 7] It is a diagram showing an example of feature data. [Figure 8] It is a diagram showing an example of a method for generating a trained model. [Figure 9] It is a diagram showing an example of the method of the first embodiment. [Figure 10] This is a diagram continuing from Figure 9. [Figure 11] It is a diagram showing an example of the method of the second embodiment. [Figure 12] It is a diagram showing an example of the method of the third embodiment. [Mode for Carrying Out the Invention]

[0012] Embodiments of the present invention will be described below with reference to the drawings. In this specification and in each drawing, elements similar to those described above in relation to previously shown drawings will be denoted by the same reference numerals, and detailed descriptions may be omitted as appropriate.

[0013] [System Overview] Figure 1 shows an example configuration of system 100 according to this embodiment. System 100 comprises a terminal 2 and a server 1 capable of communicating with terminal 2. Server 1 and terminal 2 are connected to a communication network NT such as the Internet, and are able to communicate with each other via the network. Terminal 2 is, for example, a smartphone or a tablet computer.

[0014] The support system 100 determines the user UZ's mode of transportation (MT) and visualizes the amount of carbon decarbonization when using that mode of transportation (MT). This system aims to make the user UZ aware of the impact their travel has on carbon decarbonization and encourage behavioral change in the user UZ.

[0015] The mode of transportation (MT) is, for example, a bicycle. However, it is not limited to this; the mode of transportation (MT) may also be walking, running, taking a bus, a train, or a car. Decarbonization amount refers to the amount of reduction in emissions of greenhouse gases such as carbon dioxide.

[0016] Figure 2 is a block diagram showing an example configuration of the support device 1. The support device 1 is a computer comprising a processing unit 11, a storage unit 12, and a communication unit 13.

[0017] The processing unit 11 includes a CPU and performs information processing according to the program stored in the memory unit 12. The memory unit 12 includes RAM, ROM, and non-volatile memory, etc. The communication unit 13 is an interface that connects the support device 1 to the communication network NT.

[0018] The program may be provided to the support device 1 via a non-temporary storage medium or via a communication line.

[0019] Figure 3 is a block diagram showing an example configuration of terminal 2. Terminal 2, like support device 1, is a computer equipped with a processing unit 21, a storage unit 22, and a communication unit 23. Terminal 2 further includes a touch panel 24, a camera 25, an acceleration sensor 26, and a GNSS receiver 27.

[0020] The communication unit 23 includes a communication unit that implements a mobile communication or wireless LAN communication method for connecting terminal 2 to the communication network NT, as well as a communication unit that implements short-range wireless communication such as Bluetooth®.

[0021] The touch panel 24 integrates the display unit and the operation unit. The accelerometer 26 detects the acceleration of the terminal 2. The detected acceleration is used to measure the number of steps. The GNSS receiver 27 detects the position of the terminal 2 based on radio waves received from the GNSS (Global Navigation Satellite System).

[0022] When the application program is launched, terminal 2 accumulates a history of steps and location in the background and determines the user UZ's mode of transportation MT at a predetermined time.

[0023] Incidentally, conventional methods have involved calculating the user's average speed and comparing it to a threshold to determine the user's mode of transportation. However, this method sometimes lacks sufficient accuracy in determining the user's mode of transportation.

[0024] In particular, with bicycles, the storage location of the device varies from person to person—it might be placed in a pants pocket, a mobile bag, or the bicycle basket—making it difficult to determine if it's a bicycle using conventional methods. Furthermore, factors such as speed and whether or not electric assist is used also vary from person to person, making determination difficult.

[0025] Therefore, in this embodiment, as described below, the accuracy of the determination is improved by estimating the user UZ's means of transportation MT using a pre-trained model.

[0026] [Device Functionality Configuration] Figure 4 is a block diagram showing an example of the functional configuration of terminal 2. The processing unit 21 of terminal 2 includes a step count acquisition unit 31, a location acquisition unit 32, a pre-processing unit 33, and an estimation unit 34. These functional units 31 to 34 are realized by the processing unit 21 executing information processing according to a program.

[0027] Furthermore, the processing unit 21 has access to the model storage unit 39. The model storage unit 39 may be built on the storage unit 22 or on an external storage device. The model storage unit 39 stores a trained model for estimating the means of transport MT.

[0028] The step count acquisition unit 31 acquires step count data based on the acceleration detected by the acceleration sensor 26. As shown in Figure 5, the step count data represents the number of steps taken by the user UZ per unit of time. Time represents the length of time, that is, the interval between two points in time.

[0029] The position acquisition unit 32 acquires position data based on the position detected by the GNSS receiver 27. As shown in Figure 6, the position data represents the position of terminal 2 at each point in time. A point in time represents a single point in the flow of time. The position is expressed in latitude and longitude.

[0030] The preprocessing unit 33 calculates feature data based on the step count data acquired by the step count acquisition unit 31 and the position data acquired by the position acquisition unit 32.

[0031] Figure 7 shows an example of feature data. Feature data includes features related to changes in location at each point in time. Specifically, these features related to changes in location are distance traveled, speed of travel, or time of travel. Feature data includes one or more of these features.

[0032] A change in position is defined as a change in position at a certain point in time t. n Position L n And then, at the next point in time t n+1 Position L n+1is a change between them (n is a natural number). The change between positions at a certain time point t n and the next time point t n+1 can also be rephrased as a change over the time between them. This time interval may be the same as the time interval for position detection, or may be longer than that.

[0033] The movement distance between positions is the distance at a certain time point t n between the position L n at that time point and the position L at the next time point t n+1 at that next time point n+1 The movement speed between positions is the speed obtained by dividing the movement distance between positions by the time t n+1 -t n

[0034] The movement time between positions is the time between a certain time point t n and the next time point t n+1 . Positions are basically detected at regular time intervals, but since radio waves from GNSS may not be receivable, the movement time between positions is not always constant.

[0035] The feature data further includes step count data converted to represent the number of steps between positions. That is, the step count data is converted to represent the number of steps between positions, similar to the movement distance and the like. The number of steps between positions can also be regarded as a feature quantity related to the change between positions.

[0036] Specifically, the converted step count data represents the number of steps required to walk between the position L at a certain time point t n at that time point n and the position L at the next time point t n+1 at that next time point n+1 , in other words, represents the number of steps measured during the time between a certain time point t n and the next time point t n+1 .

[0037] Note that since the step count data is originally data representing the number of steps per unit time, it may be used as-is without conversion.

[0038] ​The estimation unit 34 uses the trained model stored in the model storage unit 39 to estimate the user UZ's means of transportation MT based on the feature data (including converted step count data) calculated by the preprocessing unit 33.

[0039] [How to generate a pre-trained model] The following describes how to generate a pre-trained model. A pre-trained model is a machine learning model that has been generated using machine learning, with training feature data as input data and labels representing modes of transportation as training data.

[0040] As shown in Figure 8, in the method for generating a trained model, step count data and location data of the traveler are obtained (S61, S62), feature data is calculated (S63), labels representing the traveler's mode of transportation are assigned to the feature data (S64), and a trained model is generated by machine learning using the feature data as input data and the labels as training data (S65).

[0041] The step count data, location data, and feature data used in machine learning have the same format as the step count data, location data, and feature data described above (see Figures 5 to 7). Derived data, as described later, may also be used in machine learning.

[0042] The step count and location data of the travelers will be based on data collected when the travelers actually moved in the past. However, the step count and location data of the travelers may also be based on data generated through simulation.

[0043] Labels representing modes of transportation may include, for example, bicycles. Bicycles may be further subdivided by whether or not they are electric-assisted. Labels representing modes of transportation may also further include, for example, walking, running, buses, trains, or automobiles. Buses refer to route buses. Trains may be subdivided into electric trains, subways, and trams.

[0044] For example, a random forest can be used as a machine learning model. A random forest is a model generated by preparing multiple decision trees that represent the relationship between feature data and labels, and combining them. However, other supervised learning models, such as neural networks, may also be used as machine learning models.

[0045] In this embodiment, the accuracy of determining the user UZ's means of movement MT is improved by using feature data representing microscopic changes in the user UZ's movement, including the distance traveled between locations at each point in time, travel speed, travel time, and number of steps, as input data for a trained model. Here, "microscopic" refers to focusing on the features between individual locations rather than considering the user's movement as a whole, as in conventional methods.

[0046] For example, even when using the same mode of transportation (MT), the characteristics of changes in travel distance and speed between locations differ between urban roads with many traffic lights and wide suburban roads. By using feature data that represents these characteristics of changes, it becomes possible to accurately determine the mode of transportation (MT).

[0047] In particular, when the mode of transportation (MT) is a bicycle, differences in step count occur depending on whether terminal 2 is stored in a pants pocket, a portable bag, or the bicycle basket. Therefore, including step count data makes it possible to accurately determine whether a bicycle is being used. Furthermore, since differences in step count and travel speed occur depending on whether or not an electric assist is used, it becomes possible to determine not only whether a bicycle is being used, but also whether or not an electric assist is being used.

[0048] Therefore, in this embodiment, step count data is used to estimate means of transportation other than walking (specifically, to estimate whether or not a bicycle is used).

[0049] [First Embodiment] Figures 9 and 10 are flowcharts showing an example of the procedure of the first embodiment implemented in the support system 100. The processing unit 21 of terminal 2 executes processes S11 to S23 shown in Figures 9 and 10 according to the program. The processing unit 11 of support device 1 executes processes S31 to S34 shown in Figure 10 according to the program.

[0050] Terminal 2 monitors speed while accumulating step count and location history in the background, and groups the user UZ's movements into actions according to speed. An action refers to a segment of movement. Then, at a predetermined timing, Terminal 2 starts the process shown in Figure 9 to determine the means of movement MT for each action.

[0051] The timing for initiating processing may include, for example, when the application program becomes active, when user UZ input is received, when the completion of an action is detected, or when a predetermined time has elapsed.

[0052] As shown in Figure 9, first, terminal 2 acquires step count data (S11, processing as step count acquisition unit 31, see Figure 5). Specifically, terminal 2 extracts step count data related to the target action from the accumulated step count data.

[0053] Next, terminal 2 acquires location data (S12, processing as location acquisition unit 32, see Figure 6). Specifically, terminal 2 extracts location data related to the target action from the accumulated location data.

[0054] Next, terminal 2 calculates feature data based on the acquired step count data and location data (S13, processing as preprocessing unit 33, see Figure 7).

[0055] Next, terminal 2 determines whether the average speed of the target action is above the upper limit of the low-speed range or below the lower limit of the high-speed range (S14, S15). In other words, terminal 2 determines whether the average speed of the target action falls within the medium-speed range.

[0056] The upper limit of the low-speed range is a threshold used to distinguish walking from other faster modes of transportation. The lower limit of the high-speed range is a higher threshold than the upper limit of the low-speed range and is used to distinguish high-speed vehicles such as trains, automobiles, or airplanes from other slower modes of transportation.

[0057] If the average speed of the target action falls within the low-speed range (S14:NO), terminal 2 estimates the means of movement MT for the target action to be "walking" (S16).

[0058] If the average speed of the target action falls within the high-speed range (S15:NO), terminal 2 estimates the means of transportation MT for the target action to be a "high-speed moving object" (i.e., a high-speed moving object such as a train, automobile, or airplane) (S18).

[0059] In this embodiment, "walking" and "high-speed moving object" are determined using only the average speed, but the definition is not limited to this, and other conditions such as maximum speed may also be used in addition to the average speed to determine whether an object is "walking" or a "high-speed moving object".

[0060] If the average speed of the target action falls within the medium speed range (S14, S15: YES), terminal 2 uses the trained model to estimate the user UZ's means of transportation MT from the feature data (S17, processing as estimation unit 34).

[0061] Terminal 2 estimates whether the mode of transportation MT is a "bicycle". In addition to determining whether it is a "bicycle", it may also estimate whether or not it is electric-assisted. Terminal 2 may also estimate whether the mode of transportation MT is "running".

[0062] Furthermore, terminal 2 may estimate whether the mode of transport MT is a "vehicle". Vehicles are generally included in the high-speed range, but some vehicles such as buses may be included in the medium-speed range, so vehicles are also estimated in the medium-speed range.

[0063] The trained model is a machine learning model designed to estimate the mode of transport (MT) in the medium speed range. It outputs the probability of "bicycle," "running," and "vehicle," and the class with the highest probability is determined to be the mode of transport (MT).

[0064] As shown in Figure 10, terminal 2 then calculates the amount of carbon decarbonization for actions that qualify as decarbonization actions (S21). Specifically, terminal 2 determines that actions such as "walking," "cycling," or "running" are decarbonization actions for which the straight-line distance from the starting point to the destination is greater than or equal to a predetermined distance, and calculates the amount of carbon decarbonization for those actions. The calculation of the amount of carbon decarbonization may also be performed by server 1.

[0065] Next, terminal 2 transmits movement history data representing the user UZ's movement history to server 1 (S22), and displays the user UZ's movement history and decarbonization amount on the screen (S23). The movement history data includes data for each action, and each piece of data includes the distance traveled, the means of transportation, and the amount of decarbonization for that action.

[0066] When movement history data is transmitted from terminal 2, server 1 retrieves the transmitted movement history data and stores it in association with the user UZ identification data (S31, S32).

[0067] Next, Server 1 registers decarbonization points corresponding to the amount of decarbonization included in the movement history data, associating them with the user UZ's identification data (S33), and notifies Terminal 2 of the registered decarbonization points (S34). The amount of decarbonization and decarbonization points are converted, for example, at a conversion rate of 1g:1 point.

[0068] When Server 1 notifies Terminal 2 of a decarbonization point, Terminal 2 displays the notified decarbonization point on its screen (S24). This completes the series of processes.

[0069] According to this embodiment, since feature data representing microscopic changes in the user UZ's movement is used as described above, it becomes possible to accurately determine whether or not the user UZ is using a means of transportation MT, particularly a bicycle. Consequently, it also becomes possible to accurately calculate the amount of decarbonization.

[0070] Furthermore, by determining the user's mode of transportation (MT) and visualizing the amount of carbon decarbonization achieved when using that mode of transportation (MT), it becomes possible to make users aware of the impact their travel has on carbon decarbonization and encourage behavioral changes in them.

[0071] Furthermore, by awarding decarbonization points to users (UZ) based on their decarbonization efforts and allowing them to exchange these points for rewards offered within communities such as local governments, it becomes possible to increase the motivation of users (UZ) to accumulate decarbonization points and further encourage behavioral change among them.

[0072] [Second Embodiment] Figure 11 is a flowchart showing an example of the procedure of the second embodiment. Steps that overlap with the above embodiment may be given the same number, and detailed explanations may be omitted.

[0073] Terminal 2 calculates feature data (S13), then calculates derived data using one or more of the step count data, location data, and feature data (S41, processing as preprocessing unit 33), and uses a trained model to estimate the user UZ's means of transportation MT from the feature data and derived data (S42, processing as estimation unit 34).

[0074] The trained model is a machine learning model that estimates various modes of transportation (MT) across all speed ranges based on feature data and derived data. The trained model outputs the probability of each mode of transportation, such as "walking," "running," "cycling," "bus," "train," and "car," and the class with the highest probability is determined to be the mode of transportation (MT).

[0075] Derived data includes data that represents the continuity of features. For example, derived data may represent the continuity of movement speed calculated based on movement speed between multiple consecutive locations, or data that represents the continuity of the number of steps calculated based on the number of steps between multiple consecutive locations, or data that represents the continuity or linearity of a path calculated based on the position at multiple consecutive points in time.

[0076] For example, trains tend to have relatively high continuity of speed, while buses tend to have relatively low continuity of speed, or in other words, a relatively high frequency of stops. Running also tends to have relatively high continuity of steps. Furthermore, trains tend to follow relatively straight routes, while bicycles or cars tend to follow relatively low straight routes, or in other words, a relatively high frequency of turns. Derived data illustrates these differences in trends.

[0077] Derived data focuses on the continuity of features and represents the differences in trends described above. By using derived data along with feature data as input data to a trained model, it becomes possible to improve the accuracy of determining various modes of transport (MT) across all speed ranges.

[0078] In addition to making decisions based on a pre-trained model, the accuracy of the class may also be adjusted by comparing the travel path with map data. For example, by adjusting the accuracy of trains to be higher when the travel path follows a railway line on the map, the accuracy of determining the mode of transport (MT) can be further improved.

[0079] [Third Embodiment] Figure 12 is a flowchart showing an example of the procedure of the third embodiment. Steps that overlap with the above embodiments may be given the same number, and detailed explanations may be omitted.

[0080] If the average speed of the target action falls within the medium speed range (S15: YES), terminal 2 uses a pre-trained model for the medium speed range to estimate the user UZ's means of transportation MT from the feature data and derived data (S51, processing as estimation unit 34).

[0081] Furthermore, if the average speed of the target action falls within the high-speed range (S15:NO), terminal 2 uses a pre-trained model for the high-speed range to estimate the user UZ's means of transportation MT from the feature data and derived data (S52, processing as estimation unit 34).

[0082] The pre-trained model for the medium speed range is a machine learning model designed to estimate the mode of transportation (MT) in the medium speed range. For example, it outputs the probability of "bicycle" and "running," and the class with the highest probability is determined to be the mode of transportation (MT) in the medium speed range.

[0083] The high-speed pre-trained model is a machine learning model designed to estimate the mode of transport (MT) at high speeds. It outputs the probability of "bus," "train," and "car," and the class with the highest probability is determined to be the mode of transport (MT) at high speeds.

[0084] In this embodiment, by using a pre-trained model for the medium speed range and a pre-trained model for the high speed range depending on the average speed of the target action, it is possible to improve the determination accuracy of the means of transport MT in the medium speed range and the determination accuracy of the means of transport MT in the high speed range, respectively.

[0085] Although embodiments of the present invention have been described above, the present invention is not limited to the embodiments described above, and various modifications are of course possible for those skilled in the art.

[0086] The following lists representative embodiments of the present invention.

[0087] In the first embodiment, the program causes a computer to perform the following actions: acquire step count data representing the number of steps taken by a user holding a terminal per hour; acquire location data representing the location of the terminal at each point in time; calculate feature data representing feature quantities related to the change between the locations at each point in time based on the location data; and use a trained model generated by machine learning, with training step count data representing the number of steps per hour and training feature data representing feature quantities related to the change between locations at each point in time as input data, and labels representing the means of transportation as training data, to estimate the user's means of transportation based on the step count data and the feature data. This makes it possible to improve the accuracy of determining the user's means of transportation.

[0088] In a second embodiment, in the first embodiment, the feature quantity may include the distance traveled between the positions. This makes it possible to improve the accuracy of determining the movement means by using the distance traveled between the positions.

[0089] In a third embodiment, in the first embodiment, the feature quantity may include the speed of movement between the positions. This makes it possible to improve the accuracy of determining the movement means by using the speed of movement between positions.

[0090] In a fourth embodiment, in the first embodiment, the feature quantity may include the travel time between the positions. This makes it possible to improve the accuracy of determining the travel means by using the travel time between positions.

[0091] In a fifth embodiment, the computer may be further made to convert the step count data in the first embodiment to represent the number of steps between the locations, and the estimation of the means of movement may be performed based on the converted step count data. This makes it possible to improve the accuracy of determining the means of movement using the number of steps between locations.

[0092] In a sixth embodiment, in the first embodiment, the estimation of the means of transportation may include means of transportation other than walking. This makes it possible to improve the accuracy of determining means of transportation other than walking using step count data.

[0093] In a seventh embodiment, in the first embodiment, the estimation of the means of transportation may be an estimation of whether or not a bicycle is used. This makes it possible to improve the accuracy of determining whether or not a bicycle is used.

[0094] In the eighth aspect, in the first aspect, the estimation of the means of movement may be performed when the average speed based on the position data falls within a predetermined range. This makes it possible to limit the range of the average speed and improve the accuracy of determining the means of movement.

[0095] In a ninth aspect, in the first aspect, estimation of the means of transport may be performed using the trained model generated for the first range when the average speed based on the position data falls within the first range, and using the trained model generated for the second range when the average speed falls within a second range different from the first range. This makes it possible to improve the accuracy of determining the means of transport depending on the range of the average speed.

[0096] In a tenth embodiment, the computer may be further made to calculate derived data using one or more of the step count data, location data, and feature data in the first embodiment, and the estimation of the means of transportation may be further performed based on the derived data. This makes it possible to improve the accuracy of determining the means of transportation by using the derived data.

[0097] In an eleventh embodiment, the estimation of the means of transportation may include estimating whether or not two or more of the following are used: running, cycling, bus, train, and automobile. This makes it possible to improve the accuracy of determining multiple means of transportation.

[0098] In the twelfth embodiment, the terminal is a terminal held by a user and comprises: a step count acquisition unit that acquires step count data representing the user's step count per hour; a location acquisition unit that acquires location data representing the terminal's location at each point in time; a pre-processing unit that calculates feature data representing feature quantities related to changes between locations at each point in time based on the location data; and an estimation unit that takes learning step count data representing the step count per hour and learning feature data representing feature quantities related to changes between locations at each point in time as input data, and uses a trained model generated by machine learning as training data to estimate the user's means of transportation based on the step count data and the feature data. This makes it possible to improve the accuracy of determining the user's means of transportation.

[0099] In a 13th embodiment, the method acquires step count data representing the number of steps taken by a user holding a terminal per hour, acquires location data representing the location of the terminal at each point in time, calculates feature data representing the change in location between each point in time based on the location data, uses training step count data representing the number of steps per hour and training feature data representing the change in location between each point in time as input data, and uses labels representing the means of transportation as training data, to estimate the user's means of transportation based on the step count data and the feature data using a trained model generated by machine learning. This makes it possible to improve the accuracy of determining the user's means of transportation.

[0100] In the 14th embodiment, the method for generating a trained model involves acquiring step count data representing the number of steps a person takes per hour, acquiring location data representing the person's position at each point in time, calculating feature data representing features related to the change between positions at each point in time based on the location data, assigning labels representing the person's means of transportation to the step count data and the feature data, and generating a trained model by machine learning using the step count data and the feature data as input data and the labels representing the means of transportation as training data. This makes it possible to improve the accuracy of determining the user's means of transportation. [Explanation of Symbols]

[0101] 1 Server, 2 Terminal, 11 Processing Unit, 12 Memory Unit, 13 Communication Unit, 21 Processing Unit, 22 Memory Unit, 23 Communication Unit, 24 Touch Panel, 25 Camera, 26 Accelerometer, 27 GNSS Receiver, 31 Step Count Data Acquisition Unit, 32 Location Data Acquisition Unit, 33 Preprocessing Unit, 34 Estimation Unit, 35 Conversion Unit, 36 Speed ​​Determination Unit, 37 Derived Data Generation Unit, 41 Trained Model Storage Unit, 100 System, NT Communication Network

Claims

1. To obtain step count data representing the number of steps taken per hour by the user who possesses the device. To obtain location data representing the position of the aforementioned terminal at each point in time, Based on the aforementioned location data, feature data representing feature quantities related to the change between locations at each time point is calculated, and The system uses a trained model generated by machine learning to estimate the user's mode of transportation based on the step count data and the feature data, with training step count data representing the number of steps taken per hour and training feature data representing the change in location at each point in time as input data, and labels representing the mode of transportation as training data. A program that causes a computer to execute something.

2. The aforementioned feature includes the distance traveled between the positions. The program according to claim 1.

3. The aforementioned feature includes the speed of movement between the positions. The program according to claim 1.

4. The aforementioned feature includes the time of movement between the positions. The program according to claim 1.

5. The computer is further made to convert the aforementioned step count data so that it represents the number of steps between the aforementioned locations. The estimation of the means of transportation is performed based on the converted step count data. The program according to claim 1.

6. The estimation of the means of locomotion involves estimating means of locomotion other than walking. The program according to claim 1.

7. The estimation of the means of transportation involves estimating whether or not a bicycle is used. The program according to claim 1.

8. The estimation of the means of movement is performed when the average speed based on the position data falls within a predetermined range. The program according to claim 1.

9. The estimation of the means of transport is, When the average velocity based on the position data falls within a first range, the process is performed using the trained model generated for the first range. If the average speed falls within a second range different from the first range, the trained model generated for the second range is used. The program according to claim 1.

10. Further, the computer is made to calculate derived data using one or more of the step count data, location data, and feature data. The estimation of the means of transport is further performed based on the derived data. The program according to claim 1.

11. The estimation of the means of transportation involves estimating whether two or more of the following are used: running, cycling, bus, train, and automobile. The program according to claim 1.

12. A device owned by the user, A step count acquisition unit that acquires step count data representing the user's step count per hour, A location acquisition unit that acquires location data representing the position of the terminal at each point in time, A preprocessing unit calculates feature data representing feature quantities related to the change between positions at each time point, based on the aforementioned position data. An estimation unit takes learning step count data representing the number of steps taken per hour and learning feature data representing features related to changes in location at each point in time as input data, and uses a trained model generated by machine learning to estimate the user's means of transportation based on the step count data and the feature data, with labels representing the means of transportation as training data. A terminal equipped with the following features.

13. We obtain step count data representing the number of steps taken per hour by the user who possesses the device. Location data representing the terminal's position at each point in time is acquired. Based on the aforementioned position data, feature data representing the feature quantities related to the change between positions at each time point is calculated. Using training step count data representing the number of steps taken per hour and training feature data representing features related to changes in location at each point in time as input data, and labels representing the mode of transportation as training data, a trained model generated by machine learning is used to estimate the user's mode of transportation based on the step count data and the feature data. method.

14. We obtain step count data representing the number of steps taken by a person per hour. Location data representing the position of the aforementioned mover at each point in time is obtained. Based on the aforementioned position data, feature data representing the feature quantities related to the change between positions at each time point is calculated. Labels representing the means of transportation of the person are assigned to the step count data and the feature data. Using the aforementioned step count data and feature data as input data, and the labels representing the means of transportation as training data, a trained model is generated by machine learning. Method for generating a pre-trained model.

Images