JP2020177354A - Prediction device, prediction system, computer program, prediction method, and generation method - Google Patents

Abstract

To provide a prediction device capable of predicting a ride rate in accordance with successively changing congestion conditions.SOLUTION: The prediction device is provided with an acquisition unit for acquiring a boarding and alighting location and a boarding and alighting time of a vehicle, and a prediction unit for inputting a currently obtainable boarding and alighting rate into a learned model that has learned a past boarding and alighting rate situation and predicting the boarding rate at the boarding and alighting location at the boarding and alighting time.SELECTED DRAWING: Figure 2

Description

The present invention relates to a prediction device or the like that predicts the occupancy rate of a vehicle.

A system has been proposed that provides users of public transportation such as trains with the congestion status of vehicles such as trains. For example, the prediction system described in Patent Document 1 acquires the boarding rate from a congestion sensor at a station before the station scheduled to get off, updates the boarding rate increase / decrease table for the day, and refers to the updated boarding rate increase / decrease table. The congestion situation of the next train is predicted.

Japanese Unexamined Patent Publication No. 2013-73396

In the conventional prediction system, the congestion situation is predicted on a rule basis, so if there is an increase or decrease in the occupancy rate that is not in the rule, the congestion situation may not be predicted accurately. The present invention has been made in view of such circumstances. The purpose is to provide a predictor and the like that can predict the occupancy rate according to the gradually changing congestion situation even if the occupancy rate does not meet the rules.

The prediction device according to the present invention inputs the currently available boarding rate status into the acquisition unit that acquires the boarding / alighting place and the boarding / alighting time of the vehicle and the learned model that has learned the past boarding rate situation, and the boarding / alighting place. It is characterized by including a prediction unit for predicting the boarding rate at the boarding / alighting time.

In the present invention, the occupancy rate can be predicted according to the gradually changing congestion situation.

It is explanatory drawing which shows the configuration example of the information provision system. It is a block diagram which shows the hardware configuration example of an information providing apparatus. It is explanatory drawing which shows the example of the equipment installed in a station. It is a block diagram which shows the hardware configuration example of a mobile terminal. It is explanatory drawing which shows the example of the occupancy rate DB. It is explanatory drawing about a occupancy rate prediction model. It is a flowchart which shows the procedure example of the predicted boarding rate provision processing. It is explanatory drawing about the occupancy rate prediction model example. It is explanatory drawing which shows the display example of the predicted boarding rate. It is explanatory drawing about another example of a occupancy rate prediction model. It is explanatory drawing about another example of a occupancy rate prediction model. It is explanatory drawing which shows the other display example of the predicted occupancy rate. It is a flowchart which shows the procedure example of the learning process of the 1st prediction model. It is a flowchart which shows the procedure example of the learning process of the 2nd prediction model. It is explanatory drawing about another example of a occupancy rate prediction model.

Hereinafter, embodiments of the present invention will be described with reference to the drawings. In the following embodiment, an example applied to a railway (vehicle), which is an example of public transportation, will be described. It can also be applied to transportation other than railways, subways, buses, monorails, new transportation systems, trams, cable cars, ships, etc. In the following explanation, the station corresponds to the boarding / alighting place. The boarding / alighting place corresponds to a bus stop for a bus and a port for a ship.

FIG. 1 is an explanatory diagram showing a configuration example of an information providing system. The information providing system 100 includes an information providing device 1 (prediction device), a station 2, a vehicle 3, and a mobile terminal 4 (communication terminal). The information providing device 1, the station 2, and the vehicle 3 are communicably connected to each other via the network PN. The network PN is a dedicated line network or VPN (Virtual Private Network). Further, the information providing device 1 and the mobile terminal 4 are connected to each other so as to be able to communicate with each other via the network N. The network N is the Internet, a public telephone network, a mobile telephone network, or a communication network in which these are arbitrarily combined.

A motion sensor 31 is attached to the vehicle 3. The motion sensor 31 is a camera, a distance sensor, an infrared sensor, a load sensor, or the like. The motion sensor 31 obtains the occupancy rate from, for example, an image acquired by a camera. The motion sensor 31 obtains the occupancy rate from the values measured by the radar, the distance sensor, or the infrared sensor. The occupancy rate may be obtained from the weight of a person measured by a load sensor or a load-bearing device. Further, the occupancy rate may be obtained by using a plurality of sensors. In the case of operating a train consisting of a plurality of cars as one train on a railroad or the like, the multiple cars constituting the train are simply collectively referred to as rolling stock. In this case, the occupancy rate is the average value, maximum value, or the like of the occupancy rate of each vehicle constituting the formation.

FIG. 2 is a block diagram showing a hardware configuration example of the information providing device. The information providing device 1 is composed of a server computer, a blade server, and the like. The information providing device 1 includes a control unit 11, a main storage unit 12, an auxiliary storage unit (storage unit) 13, a communication unit 15, and a reading unit 16. The control unit 11 has one or more CPUs (Central Processing Units), MPUs (Micro-Processing Units), GPUs (Graphics Processing Units), and other arithmetic processing devices. The control unit 11 works as each functional unit such as an acquisition unit, a prediction unit, an output unit, a first output unit, and a time prediction unit by reading and executing the control program 1P stored in the auxiliary storage unit 13. The control unit 11 performs various information processing, control processing, and the like related to the information providing device 1.

The main storage unit 12 is a SRAM (Static Random Access Memory), a DRAM (Dynamic Random Access Memory), a flash memory, or the like. The main storage unit 12 temporarily stores data necessary for the control unit 11 to execute arithmetic processing.

The auxiliary storage unit 13 is a hard disk, SSD (Solid State Drive), or the like, and is a control program 1P, a first prediction model 141, a second prediction model 142, and a third prediction model 143 necessary for the control unit 11 to execute processing. And various DBs such as a boarding rate DB (DataBase: database) 131 are stored.

The communication unit 15 is a communication module for performing processing related to communication, and transmits / receives information to / from a mobile terminal 4 or the like. The reading unit 16 reads a portable storage medium 1a including a CD (Compact Disc) -ROM and a DVD (Digital Versaille Disc) -ROM. The control unit 11 may read the control program 1P from the portable storage medium 1a via the reading unit 16 and store it in the auxiliary storage unit 13. Further, the control unit 11 may download the control program 1P from another computer via the network N or the like and store it in the auxiliary storage unit 13. Furthermore, the control unit 11 may read the control program 1P from the semiconductor memory 1b.

The auxiliary storage unit 13 may be an external storage device connected to the information providing device 1. Various DBs stored in the auxiliary storage unit 13 may be stored in the cloud storage. Further, the information providing device 1 may be a multi-computer composed of a plurality of computers, a virtual machine virtually constructed by software, or a quantum computer.

FIG. 3 is an explanatory diagram showing an example of equipment installed at a station. An automatic ticket gate 21, a surveillance camera 22, a beacon 23, and the like are installed at station 2. The automatic ticket gate 21 counts the number of passengers passing by, and obtains the train occupancy rate from the counted value. When the train arrives at the platform, the surveillance camera 22 photographs the state of the platform and the inside of the train, and obtains the boarding rate based on the captured image. A person placed on the platform may visually determine the occupancy rate. The boarding rate is transmitted to the information providing device 1 via the network N.

FIG. 4 is a block diagram showing a hardware configuration example of the mobile terminal. The mobile terminal 4 is a terminal used by the user. The user is a user of transportation and is provided with a predicted occupancy rate of the train. The mobile terminal 4 is composed of a tablet computer, a smartphone and the like. The mobile terminal 4 includes a control unit 41, a main storage unit 42, an auxiliary storage unit 43, an input unit 44, a display unit 45, and a communication unit 46. Each configuration is connected by bus B.

The control unit 41 has one or more CPUs or arithmetic processing devices such as an MPU (Micro Processing Unit). Each part of the hardware is controlled according to the control program 4P stored in the auxiliary storage unit 43. By executing the control program 4P, the control unit 41 functions as each functional unit such as a reception unit, a second output unit, and a display unit. The main storage unit 42 is, for example, SRAM, DRAM or flash memory. The main storage unit 42 temporarily stores data generated when the program is executed by the control unit 41. The auxiliary storage unit 43 is a hard disk, SSD, or the like, and stores the control program 4P or the like necessary for the control unit 41 to execute the process.

The input unit 44 is, for example, a touch panel. The display unit 45 includes a liquid crystal display panel and the like. The input unit 44 and the display unit 45 form an integrated touch panel display. The input unit 44 may be a mouse or a keyboard. Further, the mobile terminal 4 may display on an external display device.

The communication unit 46 communicates with the information providing device 1 via the network N. Further, the control unit 41 may use the communication unit 46 to download the control program 4P from another computer via the network N or the like and store it in the auxiliary storage unit 43.

FIG. 5 is an explanatory diagram showing an example of a occupancy rate DB. The boarding rate DB131 stores the boarding rate (boarding rate status) for each train and each station. The occupancy rate DB131 has a matrix structure. In the boarding rate DB131, trains are lined up in the column direction, and station names (identification information) are lined up in the row direction. The element stores the time when the train arrived at the station or the time when the train left the station and the boarding rate. For example, train 123F arrived at station A at 6:00, indicating that the occupancy rate was 80%. The occupancy rate DB 131 shown in FIG. 5 is an example. As the boarding rate DB131, a table for each station may be provided, and the time and the boarding rate may be associated and stored. Further, when a plurality of lines are connected to one station, it is desirable to provide a boarding rate DB131 for each line. The occupancy rate DB 131 may be stored in a database server or cloud storage other than the information providing device 1. The station name is used as station identification information, but the station number (station number) based on the cybernetic code or station numbering set by the Japan Railway Engineers' Association may be used.

FIG. 6 is an explanatory diagram of a occupancy rate prediction model. The occupancy rate prediction model is a neural network related to LSTM (Long-Short Term Memory). LSTM is a kind of RNN (Recurrent Neural Network), and is a neural network that inputs time series data before the prediction target time point and outputs the prediction value at the target time point.

The LSTM has an input layer, an intermediate layer, and an output layer. The input layer has a plurality of neurons that receive input of parameters at each time point in chronological order. The output layer has neurons that output predicted occupancy rates. The middle layer has neurons for calculating predicted values from the input values to each neuron in the input layer. The neurons in the middle layer are called LSTM Blocks, and by using the calculation results in the middle layer for the input values at the past time point to perform the calculation for the input value at the next time point, from the time series data up to the latest time point. Calculate the value at the next time. In FIG. 6, EOS is an abbreviation for End Of Sequence and indicates the end of time series data. The generation of the occupancy rate prediction model will be described later.

Note that FIG. 6 is a Sequence-to-Sequence model which is an example of a learning model using LSTM, and the present embodiment is not limited to this. For example, the intermediate layer is not limited to one layer, and may be two or more layers. Moreover, the number of neurons in the input layer and the output layer is not limited to three. Further, the occupancy rate prediction model is not limited to the LSTM, and may be a neural network other than the LSTM.

Next, the provision of the predicted occupancy rate using the occupancy rate prediction model will be described. FIG. 7 is a flowchart showing a procedure example of the predicted occupancy rate providing process. The control unit 41 of the mobile terminal 4 receives search conditions from the user via the input unit 44 (step S1). The search condition is a route search condition. For example, the boarding station, the getting-off station, and the scheduled getting-off time (boarding / alighting time), or the boarding station, the getting-off station, and the scheduled boarding time (the boarding / alighting time). The control unit 41 transmits the search condition to the information providing device 1 via the communication unit 46 (step S2). The control unit 11 of the information providing device 1 acquires the search condition via the communication unit 15 (step S3). The control unit 11 acquires the planned route based on the search condition (step S4). Since the planned route can be realized by a known technique, the description thereof will be omitted. The planned route includes the boarding station and boarding time (time), and the getting-off station and getting-off time (time). Since the boarding time or the getting-off time is set to the time according to the timetable by the route search, it does not necessarily match the time input by the user as the search condition. It is desirable that the boarding time is later than the time specified by the user and the disembarking time is earlier than the time specified by the user. The control unit 11 acquires the latest occupancy rate from the occupancy rate DB 131 (step S5). The control unit 11 inputs the occupancy rate prediction model based on the acquired occupancy rate (step S6). The control unit 11 acquires the predicted value of the occupancy rate from the occupancy rate prediction model (step S7). The control unit 11 transmits the predicted value to the mobile terminal 4 via the communication unit 15 (step S8). The control unit 41 of the mobile terminal 4 acquires the predicted value via the communication unit 46 (step S9). The control unit 41 displays the acquired predicted value on the display unit 45 (step S10), and ends the process.

(Example of occupancy rate prediction model, part 1)
FIG. 8 is an explanatory diagram of an example of a occupancy rate prediction model. The occupancy rate prediction model shown in FIG. 8 is called a first prediction model 141 (first learning model). The first prediction model 141 outputs a predicted value of the occupancy rate of each station after the predetermined station when the occupancy rate from the first station to the predetermined station is given as an input in the same train (same vehicle). It is a prediction model. FIG. 8 shows an example in which the occupancy rate from A station to C station is input to the first prediction model 141 for the train 123F, and the predicted value of the occupancy rate of each station from D station to F station is obtained. When the user inputs the boarding station D station, the getting-off station F station, and the scheduled boarding time of 6:10 as search conditions at around 6:08, the prediction is performed as shown in FIG.

FIG. 9 is an explanatory diagram showing a display example of the predicted occupancy rate. According to the user's search conditions, the predicted boarding rates of D station and F station when the train departing from D station at 6:10 and heading for F station are displayed. The boarding rate of station E on the way may be displayed.

The search condition for which the prediction shown in FIG. 8 is performed is not unique, and the following cases are also conceivable. This is a case where the user inputs the boarding station E station, the getting-off station F station, and the scheduled boarding time 6:13 as the search conditions at 6:09. In this case, as of 6:09 when the user searched, the train 123F is heading from C station to D station, and the actual occupancy rate at D station is not available. Therefore, the boarding rate to the C station is used as the input of the first prediction model 141, and the boarding rate after the D station is predicted. Since it is considered that the user does not need the occupancy rate of C station, the predicted occupancy rate of C station is not displayed on the mobile terminal 4 as shown in FIG.

In addition, the number of input layers and output layers in the occupancy rate prediction model changes depending on the search conditions input by the user. FIG. 10 is an explanatory diagram of another example of the occupancy rate prediction model. The occupancy rate prediction model shown in FIG. 10 is the first prediction model 141 as in FIG. 8. If the user inputs the boarding station C station, the getting-off station F station, and the scheduled boarding time 6:05 as the search conditions at 6:03, the boarding rates of A station and B station will be input, and from C station to F. The estimated boarding rate to the station is the output. There are two input layers and four output layers.

The search conditions for which the prediction shown in FIG. 10 is performed may be as follows. If the user inputs boarding station D, getting off station F station, and scheduled boarding time 6:10 as search conditions at 6:03, as of 6:03, train 123F is heading from station B to station C. There is. Therefore, since the actual value of the boarding rate at C station is not available, the boarding rate at C station is also a prediction target.

As described above, in the prediction using the first prediction model 141, it is possible to predict the occupancy rate at the station arriving after that from the actual value of the occupancy rate at each station for the same train.

(Example of occupancy rate prediction model, part 2)
FIG. 11 is an explanatory diagram of another example of the occupancy rate prediction model. The occupancy rate prediction model shown in FIG. 11 is called a second prediction model 142 (second learning model). The second prediction model 142 is a occupancy rate prediction model that outputs a predicted value of the occupancy rate of the next and subsequent trains when the occupancy rate of a train that has already arrived and departed at a certain station (same station) is given as an input. In FIG. 11, for station A, the occupancy rates of train 123F departing at 6 o'clock, train 145A departing at 6:10, and train 243A departing at 6:20 are given to the second prediction model 142 as inputs, and at 6:30. This is an example of obtaining a predicted value of the occupancy rate of the train 287F departing, the train 290A departing at 6:40, and the train 304A departing at 6:50. For example, when the user instructs a search at around 6:22. It is possible to predict the occupancy rate at stations other than A as well as at station A. The second prediction model 142 may be common to all stations, or a second prediction model 142 may be provided for each station.

FIG. 12 is an explanatory diagram showing another display example of the predicted occupancy rate. FIG. 12 shows the situation of station A as of 6:22. The actual occupancy rate is displayed in the upper table, and the predicted occupancy rate is displayed in the lower table. Further, the occupancy rate may be predicted after the time displayed in FIG. 12, and the occupancy rate that can be said to have alleviated congestion, for example, the time below 100% may be predicted and displayed. In the example of FIG. 12, it is displayed that the congestion will be alleviated from about 7:30.

When predicting the occupancy rate with the above-mentioned route search, it is as follows. Find the planned route from the search conditions. The trains included in the obtained planned route are specified, and the predicted boarding rates of the specified trains at the boarding station and the getting-off station are obtained by the second prediction model 142. The obtained predicted boarding rate is output as a route search result. By the above procedure, it is possible to predict the occupancy rate linked with the route search even in the prediction using the second prediction model 142.

Also in the second prediction model 142, when the actual data that needs to be input for prediction is insufficient, the necessary actual data is predicted first. The predicted value and the actual data are input to the second prediction model 142 to predict the desired occupancy rate. This idea is the same as in the case of the first prediction model 141.

The occupancy rate may be predicted using the first prediction model 141 and the second prediction model 142. Then, the final predicted value may be obtained from the two predicted values. The final predicted value is appropriately selected from the average value of the two values, the weighted average value, the maximum value, and the like.

(Learning of prediction model)
Next, learning of the prediction model will be described. FIG. 13 is a flowchart showing a procedure example of the learning process of the first prediction model. The control unit 11 of the information providing device 1 acquires the target period (step S21). The target period is set in advance by the administrator. The control unit 11 acquires the occupancy rate data for the target period from the occupancy rate DB 131 (step S22). The control unit 11 selects a train to be processed from the acquired data (step S23). Even if the train is the same, the data of trains with different operating dates are treated as different trains and processed separately. The control unit 11 substitutes 1 for the variable n and substitutes the number of departure / arrival stations of the train to be processed into the variable m (step S24). The number of departure and arrival stations includes the first and last stations. The control unit 11 creates learning data having n input layers and mn output layers (step S25). The control unit 11 performs learning using the created data as teacher data in the first prediction model 141 (step S26). The control unit 11 increments the variable n by 1 (step S27). The control unit 11 determines whether or not the variable n is equal to the variable m (step S28). When the control unit 11 determines that the variable n is not equal to the variable m (NO in step S28), the process returns to step S25, and learning is performed by changing the numbers of the input layer and the output layer. When the control unit 11 determines that the variable n is equal to the variable m (YES in step S28), the control unit 11 determines whether or not there is an unprocessed train among the trains to be processed (step S29). When the control unit 11 determines that there is an unprocessed train (YES in step S29), the control unit 11 returns the process to step S23 and performs the process for the unprocessed train. When the control unit 11 determines that there is no unprocessed train (NO in step S29), the control unit 11 ends the process. The prediction model is a trained model.

The first prediction model 141 performs initial learning and then retrains at predetermined intervals. The procedure for re-learning is the same as in FIG. For example, in the case of daily re-learning, after the last train arrives at the terminal station, the occupancy rate data from the first train (first train) to the last train (last train) is acquired and re-learning is performed. In this case, the target period acquired in step S21 includes the time, and the data for one day can be acquired in step S22. Alternatively, the occupancy rate data may be flagged so that it can be determined whether or not it has been used as teacher data. The initial learning and re-learning may be performed by a computer different from the information providing device 1. In this case, the first prediction model 141 after the initial learning or the re-learning is transferred to the information providing device 1. Alternatively, the information providing device 1 stores the first prediction model 141 before learning, transfers the parameters characterizing the first prediction model 141 after learning to the information providing device 1, and updates the first prediction model 141. You may.

Subsequently, the learning of the second prediction model 142 will be described. FIG. 14 is a flowchart showing a procedure example of the learning process of the second prediction model. The control unit 11 of the information providing device 1 acquires the target period (step S41). The target period is set in advance by the administrator. The control unit 11 acquires the occupancy rate data for the target period from the occupancy rate DB 131 (step S42). The control unit 11 selects a processing date (step S43). Here, the processing date is between the time when the first train departs from the first station and the time when the last train arrives at the last station. The control unit 11 selects a station to be processed (step S44). The control unit 11 substitutes 1 for the variable n and substitutes the number of trains arriving and departing at the selected station in the variable m on the processing date (step S45). The control unit 11 creates learning data having n input layers and mn output layers (step S46). The control unit 11 performs learning using the created data as teacher data in the second prediction model 142 (step S47). The control unit 11 increments the variable n by 1 (step S48). The control unit 11 determines whether or not the variable n is equal to the variable m (step S49). When the control unit 11 determines that the variable n is not equal to the variable m (NO in step S49), the process returns to step S46, and learning is performed by changing the numbers of the input layer and the output layer. When the control unit 11 determines that the variable n is equal to the variable m (YES in step S49), the control unit 11 determines whether or not there is an unprocessed station among the processing target stations (step S50). When the control unit 11 determines that there is an unprocessed station (YES in step S50), the control unit 11 returns the process to step S44 and performs the process for the unprocessed station. When the control unit 11 determines that there is no unprocessed station (NO in step S50), the control unit 11 determines whether or not there is an unprocessed day in the processing target period (step S51). When the control unit 11 determines that there is an unprocessed day (YES in step S51), the control unit 11 returns the process to step S43 and performs the process for the unprocessed day. When the control unit 11 determines that there is no unprocessed day (NO in step S51), the control unit 11 ends the process.

The second prediction model 142 performs initial learning and then retrains at predetermined intervals. The procedure for re-learning is the same as in FIG. For example, in the case of daily re-learning, after the last train arrives at the terminal station, the occupancy rate data from the first train to the last train is acquired and re-learning is performed. The initial learning and re-learning may be performed by a computer different from the information providing device 1.

(Example of occupancy rate prediction model, part 3)
FIG. 15 is an explanatory diagram of another example of the occupancy rate prediction model. The occupancy rate prediction model shown in FIG. 15 is called a third prediction model 143. The third prediction model 143 outputs the predicted value of the occupancy rate of each station up to the next train when the occupancy rate of each station up to a certain station is given as an input for the train before a certain train. It is a model. In FIG. 15, for train 123F before train 145A, if the occupancy rates of each station up to C station, that is, A station, B station and C station are given as input, each station up to C station of the next train 145A (A station, This is an example of obtaining a predicted value of the occupancy rate of stations B and C). For example, when the user instructs a route search from station A to station C at around 6:06.

As described above, by using the occupancy rate prediction model, the future occupancy rate is predicted from the actual value of the occupancy rate. As a result, the occupancy rate can be predicted according to the gradually changing congestion situation. In addition, if the actual occupancy rate for predicting the occupancy rate at the train or station requested by the user has not been obtained due to the time when the user executed the search, the occupancy rate is predicted by the prediction model. By treating the predicted occupancy rate as a result and inputting it to the prediction model, it is possible to predict the occupancy rate at the train or station desired by the user. The mobile terminal 4 may perform the process of predicting the occupancy rate. In this case, the occupancy rate DB 131, the first prediction model 141, the second prediction model 142, and the third prediction model 143 may be stored in the auxiliary storage unit 43 of the mobile terminal 4.

When predicting the occupancy rate in the information providing device 1, any of the first prediction model 141, the second prediction model 142, and the third prediction model 143 may be used. Further, the predicted value provided to the user may be calculated from the predicted values of the plurality of prediction models. For example, using two prediction models, the first prediction model 141 and the second prediction model 142, the prediction that provides the user with the maximum value, the minimum value, the average value, the weighted average value, and the like of the prediction values obtained from the respective prediction models. Let it be a value.

Retraining of the prediction model is performed at predetermined intervals, for example, daily, but is not limited to this. It may be performed every time the boarding rate DB 131 is updated. Further, when one train arrives at the terminal station, it may be relearned using the data about the train.

The technical features (constituent requirements) described in each embodiment can be combined with each other, and by combining them, new technical features can be formed.
The embodiments disclosed this time should be considered to be exemplary in all respects and not restrictive. The scope of the present invention is indicated by the scope of claims, not the above-mentioned meaning, and is intended to include all modifications within the meaning and scope equivalent to the scope of claims.

100 Information providing system 1 Information providing device 11 Control unit 12 Main storage unit 13 Auxiliary storage unit 131 Ride rate DB
141 1st prediction model 142 2nd prediction model 143 3rd prediction model 15 Communication unit 16 Reading unit 1P Control program 1a Portable storage medium 1b Semiconductor memory 4 Mobile terminal 41 Control unit 42 Main storage unit 43 Auxiliary storage unit 44 Input unit 45 Display 46 Communication 4P Control program B Bus N Network PN Network

Claims (13)

The acquisition department that acquires the boarding / alighting place and boarding / alighting time of the vehicle,
It is characterized in that it is provided with a prediction unit that inputs the currently available occupancy rate status into a learned model that has learned the past occupancy rate status and predicts the occupancy rate at the boarding / alighting time of the boarding / alighting place. Predictor. The prediction device according to claim 1, further comprising an output unit that outputs the predicted occupancy rate. The prediction device according to claim 1 or 2, further comprising a time prediction unit that predicts a time when the occupancy rate falls below a predetermined value based on the predicted occupancy rate. The prediction device according to any one of claims 1 to 3, wherein the trained model is a recursive neural network trained using time-series data of the occupancy rate in the past. The prediction device according to claim 4, wherein the time-series data of the occupancy rate includes the occupancy rate at different boarding / alighting locations of the same vehicle. The prediction device according to claim 4, wherein the time-series data of the occupancy rate includes the occupancy rate of different vehicles at the same boarding / alighting place. The prediction unit is a first learning model learned using the time-series data of the past occupancy rate including the occupancy rate at the different boarding / alighting locations of the same vehicle, and the past at different vehicles at the same station. The prediction device according to claim 4, wherein the occupancy rate is predicted based on a second learning model learned by using the time series data of the occupancy rate. The prediction device according to any one of claims 1 to 7, wherein the boarding rate is acquired by a sensor provided at the boarding / alighting place or the vehicle. The prediction device according to claim 8, wherein the sensor sequentially acquires the boarding rate, and stores the acquired boarding rate in a storage unit in association with the identification information of the boarding / alighting place. The acquisition department that acquires the boarding / alighting place and boarding / alighting time of the vehicle,
A prediction unit that inputs the currently available occupancy rate status into a trained model that has learned the past occupancy rate status and predicts the occupancy rate at the boarding / alighting time at the boarding / alighting place.
A prediction device having a first output unit that outputs the predicted boarding rate, a reception unit that accepts inputs of the boarding / alighting place and the boarding / alighting time, and
A second output unit that outputs the received boarding / alighting place and the boarding / alighting time to the prediction device, and
A prediction system including a communication terminal having a display unit for displaying the predicted occupancy rate output by the first output unit. Obtain the boarding / alighting place and boarding / alighting time of the vehicle,
The feature is that the currently available occupancy rate status is input to the trained model that has learned the past occupancy rate status, and the computer executes a process of predicting the occupancy rate at the boarding / alighting time of the boarding / alighting place. Computer program to do. A computer that can access the storage
Obtain the boarding / alighting place and boarding / alighting time of the vehicle,
A prediction method characterized by inputting the currently available occupancy rate status into a trained model that has learned the past occupancy rate status, and predicting the occupancy rate at the boarding / alighting time of the boarding / alighting place. Acquire time-series data of the boarding rate at the boarding / alighting place of the vehicle,
A trained model characterized by having a computer execute a process of generating a learning model that outputs a predicted value of the boarding rate when the acquired data of the boarding rate at the boarding / alighting place is sequentially input in chronological order. Generation method.

Images