Subway station passenger flow prediction method, device, equipment, medium and product

Abstract

The invention provides a subway station passenger flow prediction method, device and equipment, a medium and a product, and the method comprises the steps: obtaining historical outbound passenger flow data which comprises the time sequence of the outbound passenger flow number of a target subway station; the time sequence of the number of departure people of the target subway station is a sequence formed by the number of departure people of the target subway station at a plurality of preset time points in a first historical time period; encoding the historical outbound passenger flow data to obtain a target encoding time sequence; and obtaining a dependency relationship among the data in the target coding time sequence, performing feature extraction on the data in the target coding time sequence, and predicting an outbound passenger flow data prediction value of the target subway station in the future preset time based on the obtained extracted features and the dependency relationship. According to the method, the global information and the local information in the time sequence are concerned at the same time, so that the relation between the elements in the sequence is better captured, and the passenger flow information is accurately predicted.

Description

Technical Field

[0001] The embodiments of the present invention relate to the field of artificial intelligence technology, and in particular to a method, device, equipment, medium and product for predicting passenger flow at subway stations. Background Technology

[0002] In current technology, with the opening of subway lines, subway travel has become the first choice for most people. However, this increasing number of people choosing subways puts significant pressure on rail transit, frequently leading to congestion and jostling on and off trains, potentially causing accidents. Therefore, predicting the number of subway passengers to reduce congestion and ensure passenger safety is crucial. Current methods for subway passenger prediction mostly rely solely on historical exit passenger time series data, neglecting external factors. This leads to a degree of bias in the prediction results, affecting their reliability and accuracy. Summary of the Invention

[0003] This invention provides a method, apparatus, equipment, medium, and product for predicting passenger flow at subway stations, in order to solve the problem that existing methods cannot efficiently and accurately predict the number of people traveling by subway.

[0004] To solve the above-mentioned technical problems, the present invention is implemented as follows:

[0005] In a first aspect, embodiments of the present invention provide a method for predicting passenger flow at subway stations, including:

[0006] Acquire historical exit passenger flow data, which includes: a time series of the number of people exiting the target subway station; the time series of the number of people exiting the target subway station is a sequence composed of the number of people exiting the target subway station at multiple preset time points within a first historical period;

[0007] The historical outbound passenger flow data is encoded to obtain the target encoded time series;

[0008] The dependencies between data in the target encoded time series are obtained, and features are extracted from the data in the target encoded time series to obtain the extracted features. Based on the dependencies and the extracted features, the predicted value of the exit passenger flow data of the target subway station within a future preset time period is predicted.

[0009] Optionally, the historical exit passenger flow data may further include: passenger flow data of adjacent stations of the target subway station, wherein the passenger flow data of the adjacent stations includes: a sequence of the number of people exiting or entering the station within the first historical time period.

[0010] The historical exit passenger flow data is encoded to obtain a time series of the encoded exit passenger numbers, including:

[0011] The time series of exits at the target subway station and the passenger flow data of the adjacent stations are encoded respectively;

[0012] The time series of exits at the coded target subway station is integrated with the passenger flow data of the adjacent stations to obtain the target coded time series.

[0013] Optionally, obtain historical exit passenger flow data, including:

[0014] Obtain the target base station associated with the target subway station, and obtain the mobile phone signaling data of the user under the target base station during the first historical time period;

[0015] Based on the mobile phone signaling data of users under the target base station during the first historical period, target users with subway riding behavior characteristics are obtained.

[0016] Determine whether the base station the target user passes through within a preset time after arriving at the target subway station is a base station along the subway route. If the base station the target user passes through is not a base station along the subway route, then the target user is counted as one of the exit passengers at the target subway station.

[0017] The number of people exiting the target subway station during the first historical period is correlated with a preset time point to form a time series of the number of people exiting the target subway station.

[0018] Optionally, the step of obtaining target users with subway-riding behavior characteristics based on the mobile phone signaling data of users under the target base station during the first historical time period includes:

[0019] Obtain mobile phone signaling data of all users under the target base station during the first historical time period;

[0020] Behavioral feature analysis is performed on the timestamp information in the mobile phone signaling data of all users to determine the type of the user. The user types include: target users with subway riding behavior characteristics, users who pass through, and users who stay.

[0021] Optionally, the step of performing behavioral feature analysis on the timestamp information in the mobile phone signaling data of all users to determine the type of the user includes:

[0022] The user's travel trajectory is determined based on the timestamp information in the mobile phone signaling data of all users;

[0023] The user's travel trajectory is compared with the subway route. If the similarity between the user's travel trajectory and the subway route meets a preset threshold, the user is considered a target user with subway riding behavior characteristics.

[0024] Optionally, the historical outbound passenger flow data includes historical outbound passenger flow data with multiple different date characteristics, including holidays, weekdays, and weekends.

[0025] Optionally, obtaining the time series of exiting passengers at the target subway station further includes:

[0026] Obtain the time series of exit passenger numbers at at least one adjacent station of the target subway station;

[0027] The correlation between the target subway station and the adjacent stations was determined using the Pearson correlation coefficient.

[0028] Obtain the time series of outbound passenger numbers from adjacent stations that meet the preset correlation threshold;

[0029] The time series of exits at the target subway station is weighted based on the time series of exits at adjacent stations that meet the preset correlation threshold, resulting in a weighted time series of exits at the target subway station. This weighted time series of exits at the target subway station is then used as the final time series of exits at the target subway station.

[0030] Optionally, using a pre-trained subway station passenger flow prediction model, the following steps are performed: encoding the historical exit passenger flow data to obtain a target encoded time series; obtaining the dependencies between data in the target encoded time series; extracting features from the data in the target encoded time series to obtain extracted features; and predicting the exit passenger flow data of the target subway station within a preset time period based on the dependencies and extracted features.

[0031] Optionally, the subway station passenger flow prediction model includes: a feature embedding layer, an encoder layer, and a decoder layer;

[0032] The feature embedding layer encodes the historical outbound passenger flow data to obtain the target encoded time series;

[0033] The encoder layer obtains the dependencies between data in the target encoded time series, performs feature extraction on the data in the target encoded time series, and obtains the extracted features;

[0034] Based on the dependencies and extracted features, the decoder layer predicts the exit passenger flow data of the target subway station within a preset time period.

[0035] Optionally, the encoder layer includes multiple sub-encoder layers, the output of each sub-encoder layer is used as the input of the next sub-encoder layer, and the output of the last sub-encoder layer is used as the input of the decoder layer;

[0036] The decoder layer includes multiple sub-decoder layers. The output of each sub-decoder layer is used as the input of the next sub-decoder layer. The last sub-decoder layer outputs the predicted value of the exit passenger flow data of the target subway station within a preset time period.

[0037] Each sub-encoder layer includes an attention mechanism and a feedforward neural network layer; the attention mechanism interacts and correlates the time series data of the input to the sub-encoder layer to capture the dependencies between the data; the feedforward neural network layer extracts features from the output of the sub-encoder layer.

[0038] Optionally, the subway station passenger flow prediction model is trained based on a training set, which includes: historical exit passenger flow data samples and the actual values ​​of exit passenger flow data of the target subway station. The historical exit passenger flow data samples include a sequence of the number of people exiting the target subway station in a second historical period, and the actual values ​​of the exit passenger flow data of the target subway station are the exit passenger flow data of the target subway station in a third historical period, which is later than the second historical period.

[0039] Optional, also includes:

[0040] The predicted passenger flow data of the target subway station within a preset time period is analyzed to generate a passenger flow planning, guidance and scheduling scheme.

[0041] Thirdly, embodiments of the present invention provide a subway station passenger flow prediction device, comprising:

[0042] The acquisition module is used to acquire historical exit passenger flow data, which includes: a time series of the number of people exiting the target subway station; the time series of the number of people exiting the target subway station is a sequence composed of the number of people exiting the target subway station at multiple preset time points within a first historical period;

[0043] The first processing module is used to encode the historical outbound passenger flow data to obtain the target encoded time series;

[0044] The second processing module is used to obtain the dependencies between data in the target encoded time series, extract features from the data in the target encoded time series, and obtain the extracted features; based on the dependencies and the extracted features, predict the predicted value of the exit passenger flow data of the target subway station within a future preset time.

[0045] Optionally, the historical exit passenger flow data may further include: passenger flow data of adjacent stations of the target subway station, wherein the passenger flow data of the adjacent stations includes: a sequence of the number of people exiting or entering the station within the first historical time period.

[0046] The first processing module includes:

[0047] The first processing submodule is used to encode the time series of the number of people exiting the target subway station and the passenger flow data of the adjacent stations respectively; and to integrate the time series of the number of people exiting the target subway station and the passenger flow data of the adjacent stations to obtain the target encoded time series.

[0048] Optionally, the acquisition module includes:

[0049] The first acquisition submodule is used to acquire the target base station associated with the target subway station and acquire the mobile phone signaling data of the user under the target base station during the first historical time period;

[0050] The second processing submodule is used to obtain target users with subway riding behavior characteristics based on the mobile phone signaling data of users under the target base station in the first historical time period.

[0051] The third processing submodule is used to determine whether the base station that the target user passes through is a base station along the subway route within a preset time after the target user arrives at the target subway station. If the base station that the target user passes through is not a base station along the subway route, then the target user is counted as the number of people exiting the target subway station.

[0052] The fourth processing submodule is used to associate the number of people exiting the target subway station within the first historical period with a preset time point to form a time series of the number of people exiting the target subway station.

[0053] Optionally, the second processing submodule is configured to include:

[0054] The fifth processing submodule is used to acquire mobile phone signaling data of all users under the target base station within the first historical time period; perform behavioral feature analysis on the timestamp information in the mobile phone signaling data of all users to determine the type of the user, which includes: target users with subway riding behavior characteristics, users passing through, and users staying.

[0055] Optionally, the fifth processing submodule includes:

[0056] The sixth processing submodule is used to determine the user's travel trajectory based on the timestamp information in the mobile phone signaling data of all users; compare the user's travel trajectory with the subway transportation route; if the similarity between the user's travel trajectory and the subway transportation route meets a preset threshold, then the user is considered to be a target user with subway riding behavior characteristics.

[0057] Optionally, the historical outbound passenger flow data includes historical outbound passenger flow data with multiple different date characteristics, including holidays, weekdays, and weekends.

[0058] Optionally, the acquisition module may also include:

[0059] The second acquisition submodule is used to acquire the time series of the number of people exiting at at least one adjacent station of the target subway station;

[0060] The seventh processing submodule is used to determine the correlation between the target subway station and the adjacent stations using the Pearson correlation coefficient;

[0061] The third acquisition submodule is used to acquire the time series of the number of people leaving adjacent stations that meet the preset correlation threshold;

[0062] The eighth processing submodule is used to weight the time series of exits of the target subway station based on the time series of exits of adjacent stations that meet the preset correlation threshold, to obtain the weighted time series of exits of the target subway station, and to use the weighted time series of exits of the target subway station as the final determined time series of exits of the target subway station.

[0063] Optionally, using a pre-trained subway station passenger flow prediction model, the following steps are performed: encoding the historical exit passenger flow data to obtain a target encoded time series; obtaining the dependencies between data in the target encoded time series; extracting features from the data in the target encoded time series to obtain extracted features; and predicting the exit passenger flow data of the target subway station within a preset time period based on the dependencies and extracted features.

[0064] Optionally, the subway station passenger flow prediction model includes: a feature embedding layer, an encoder layer, and a decoder layer;

[0065] The feature embedding layer encodes the historical outbound passenger flow data to obtain the target encoded time series;

[0066] The encoder layer obtains the dependencies between data in the target encoded time series, performs feature extraction on the data in the target encoded time series, and obtains the extracted features;

[0067] Based on the dependencies and extracted features, the decoder layer predicts the exit passenger flow data of the target subway station within a preset time period.

[0068] Optionally, the encoder layer includes multiple sub-encoder layers, the output of each sub-encoder layer is used as the input of the next sub-encoder layer, and the output of the last sub-encoder layer is used as the input of the decoder layer;

[0069] The decoder layer includes multiple sub-decoder layers. The output of each sub-decoder layer is used as the input of the next sub-decoder layer. The last sub-decoder layer outputs the predicted value of the exit passenger flow data of the target subway station within a preset time period.

[0070] Each sub-encoder layer includes an attention mechanism and a feedforward neural network layer; the attention mechanism interacts and correlates the time series data of the input to the sub-encoder layer to capture the dependencies between the data; the feedforward neural network layer extracts features from the output of the sub-encoder layer.

[0071] Optionally, the subway station passenger flow prediction model is trained based on a training set, which includes: historical exit passenger flow data samples and the actual values ​​of exit passenger flow data of the target subway station. The historical exit passenger flow data samples include a sequence of the number of people exiting the target subway station in a second historical period, and the actual values ​​of the exit passenger flow data of the target subway station are the exit passenger flow data of the target subway station in a third historical period, which is later than the second historical period.

[0072] Optional, also includes:

[0073] The analysis module is used to analyze the predicted values ​​of passenger flow data of the target subway station within a preset time period, and generate passenger flow planning, guidance and scheduling schemes.

[0074] Thirdly, embodiments of the present invention provide an electronic device, including a processor, a memory, and a program or instructions stored in the memory and executable on the processor, wherein when the program or instructions are executed by the processor, they implement the steps in the subway station passenger flow prediction method as described in any one of the first aspects.

[0075] Fourthly, embodiments of the present invention provide a readable storage medium storing a program or instructions that, when executed by a processor, implement the steps in the subway station passenger flow prediction method as described in any one of the first aspects.

[0076] Fifthly, embodiments of the present invention provide a computer program product, including computer instructions that, when executed by a processor, implement the steps in the subway station passenger flow prediction method as described in any one of the first aspects.

[0077] In this invention, historical exit passenger flow data is acquired, including a time series of exit passenger numbers at a target subway station. This time series is a sequence of exit passenger numbers at multiple preset time points within a first historical period. The historical exit passenger flow data is encoded to obtain a target encoded time series. Dependencies between data points in the target encoded time series are obtained, and features are extracted from the data in the target encoded time series to obtain extracted features. Based on the dependencies and extracted features, the predicted exit passenger flow data for the target subway station within a preset future time period is calculated. By simultaneously focusing on both global and local information in the time series of exit passenger numbers at the target subway station, information from different locations is integrated, thereby better capturing the relationships between elements in the sequence, improving prediction accuracy, and solving the problem of existing methods that cannot efficiently and accurately predict the number of subway passengers. Attached Figure Description

[0078] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit the invention. Furthermore, the same reference numerals denote the same parts throughout the drawings. In the drawings:

[0079] Figure 1 This is a flowchart of a subway station passenger flow prediction method provided by an embodiment of the present invention;

[0080] Figure 2 This is a schematic diagram illustrating the number of people participating in various activities at different times on weekdays in a training method for a subway station passenger flow prediction model provided in an embodiment of the present invention.

[0081] Figure 3 This is a schematic diagram illustrating the number of people participating in various activities at different times on non-working days in a training method for a subway station passenger flow prediction model provided in an embodiment of the present invention.

[0082] Figure 4 This is a line graph showing the number of people passing through and staying at different times on weekdays in a training method for a subway station passenger flow prediction model provided in an embodiment of the present invention.

[0083] Figure 5This is a flowchart illustrating the overall process of obtaining the time series of exit numbers in a training method for a subway station passenger flow prediction model provided in this embodiment of the invention.

[0084] Figure 6 This is a schematic diagram of the station correlation of a training method for a subway station passenger flow prediction model provided in an embodiment of the present invention;

[0085] Figure 7 This is a schematic diagram of the overall process of training a subway station passenger flow prediction model according to an embodiment of the present invention.

[0086] Figure 8 This is a schematic diagram of the structure of a subway station passenger flow prediction device provided in an embodiment of the present invention;

[0087] Figure 9 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0088] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0089] Please refer to Figure 1 This invention provides a method for predicting passenger flow at subway stations, including:

[0090] Step 11: Obtain historical exit passenger flow data, which includes: a time series of the number of people exiting the target subway station; the time series of the number of people exiting the target subway station is a sequence composed of the number of people exiting the target subway station at multiple preset time points within a first historical period;

[0091] In this embodiment of the invention, historical exit passenger flow data can display information about users' time dimension and spatial behavioral characteristics, providing data support for subsequent model processing.

[0092] In this embodiment of the invention, the historical exit passenger flow data further includes: passenger flow data of adjacent stations of the target subway station. The passenger flow data of the adjacent stations includes: a sequence of the number of people exiting or entering the adjacent stations within a first historical period. The adjacent stations are subway stations located at train stations or airports. The sequence of the number of people exiting or entering the adjacent stations within the first historical period is obtained according to the direction of the subway station. By integrating the historical exit passenger flow data of the target subway station with the passenger flow data of adjacent stations, more comprehensive contextual information is obtained, ensuring full utilization of historical data for analysis and helping to improve the accuracy of prediction.

[0093] In this embodiment of the invention, optionally, obtaining historical outbound passenger flow data includes:

[0094] Obtain the target base station associated with the target subway station, and obtain the mobile phone signaling data of the user under the target base station during the first historical time period;

[0095] Based on the mobile phone signaling data of users under the target base station during the first historical period, target users with subway riding behavior characteristics are obtained.

[0096] Determine whether the base station the target user passes through within a preset time after arriving at the target subway station is a base station along the subway route. If the base station the target user passes through is not a base station along the subway route, then the target user is counted as one of the exit passengers at the target subway station.

[0097] The number of people exiting the target subway station during the first historical period is correlated with a preset time point to form a time series of the number of people exiting the target subway station.

[0098] In this embodiment of the invention, since the locations of each subway station are different, the time series of their exit passengers will also be affected by the surrounding environment. For some areas with a large number of people, such as large commercial districts, famous scenic spots or schools, the corresponding subway stations will have higher passenger flow, while some remote locations will have relatively less passenger flow. Therefore, by using the ARcGis tool, the spatial geolocation of each Point of Interest (POI) within the city is achieved. In order to measure the influence range of surrounding residents on subway stations, the influence range can be divided according to the distance reached within ten minutes of walking time (approximately 800~1500m). The remaining POIs within the influence range of each subway station, excluding the current subway station, are summarized, and the geographical location level is divided according to the POI information points within the range. For example, if the surrounding area of ​​the subway includes scenic spots, schools, medical institutions, high-speed rail stations and large shopping malls, the current subway station is classified as a high-passenger-flow station; while for some remote stations or undeveloped areas, the current subway station is classified as a low-passenger-flow station. Different weights can be assigned according to the location of the surrounding POI information points to represent the importance of their spatial distribution location.

[0099] Then, based on the information points, the corresponding base station information is obtained. Since most base stations have a fan-shaped radiation range of about 400 meters, the influence circle of each subway station is divided according to the distance that can be reached in ten minutes by walking, based on the information points of each subway station and the ARcGis tool. The latitude and longitude information is matched with the corresponding base station information to filter out the list of base stations within the influence range of the subway station and associate them with the corresponding stations. Based on the base station list information associated with the target subway station, the mobile phone signaling data of the corresponding users is counted as the raw data of passenger flow.

[0100] In this embodiment of the invention, optionally, obtaining target users with subway-riding behavior characteristics based on mobile phone signaling data of users under the target base station within the first historical time period includes:

[0101] Obtain mobile phone signaling data of all users under the target base station during the first historical time period;

[0102] Behavioral feature analysis is performed on the timestamp information in the mobile phone signaling data of all users to determine the type of the user. The user types include: target users with subway riding behavior characteristics, users who pass through, and users who stay.

[0103] In this embodiment of the invention, optionally, the step of performing behavioral feature analysis on the timestamp information in the mobile phone signaling data of all users to determine the type of the user includes:

[0104] The user's travel trajectory is determined based on the timestamp information in the mobile phone signaling data of all users;

[0105] The user's travel trajectory is compared with the subway route. If the similarity between the user's travel trajectory and the subway route meets a preset threshold, the user is considered a target user with subway riding behavior characteristics.

[0106] In this embodiment of the invention, behavioral feature analysis is performed on the timestamp information in the mobile phone signaling data of all users to determine the type of the user, specifically including:

[0107] Using the ARcGis tool, the latitude and longitude information of the base station list within the influence range of the target subway station is spatially geolocated, and the correspondence between the target subway station and the base station is established, such as: station 1 corresponds to base station A and base station B, station 2 corresponds to base station B, and station 3 corresponds to base station C and base station D.

[0108] Based on the order in which the subway reaches each station, the corresponding associated base stations are arranged in that order to form the base station path along the route to the station.

[0109] Filter signaling data within the site's influence area and remove missing values, invalid data, and other abnormal data;

[0110] A time feature is added to the base station path of the station to form the base station path of the station within a specific time period; the time feature is the preset subway operation information, which includes: the subway departure time, arrival time and running speed; since the subway departure time, arrival time and running speed are usually carried out according to plan, the base station path of the station within a specific time period can be determined.

[0111] Based on the user's mobile phone signaling data, a base station path is formed within a specific time period in the above-mentioned subway operation information. If the base station path of the user is the same as the base station path of the target subway station, it is considered that the target user travels by subway to the target subway station.

[0112] For the target user who is identified to travel by subway, if the base station path of the user leaves the base station path of the subway station within a certain period of time after a specific time, the user will be counted as the number of people leaving the station at the station corresponding to the base station that captured the user's signaling data before leaving the base station path of the subway station. That is, the target user who leaves the base station path after the target subway station will be counted as the number of people leaving the target subway station.

[0113] The number of people exiting each subway station is confirmed using the method described above, and the statistical number of people exiting is correlated with the corresponding time period to form a time series of the number of people exiting the target subway station.

[0114] In this embodiment of the invention, by matching information points with base station information to form a base station list, the accuracy of subsequent signaling data is ensured. Furthermore, by acquiring the mobile phone signaling data of the target user, the user's travel patterns are monitored and evaluated, and efficiently matched with subway traffic routes and preset operation information to ensure the accuracy of predictions, reduce data noise, and improve representativeness and effectiveness.

[0115] In this embodiment of the invention, optionally, the historical outbound passenger flow data includes historical outbound passenger flow data with multiple different date characteristics, including holidays, weekdays, and weekends.

[0116] In this embodiment of the invention, the date can also be decomposed into the following features based on the mobile phone signaling timestamp characteristics:

[0117] 1) Holidays: Statutory holidays such as New Year's Day, National Day, and Spring Festival;

[0118] 2) Working days: Monday to Friday, excluding public holidays and weekends;

[0119] 3) Weekends: Regular Saturdays and Sundays excluding weekdays;

[0120] Since the number of passengers varies depending on the date, the time series of passengers exiting the station corresponding to different types of dates are integrated and analyzed separately by decomposing the date characteristics, thereby improving accuracy.

[0121] In this embodiment of the invention, the classification of users includes: target users with subway riding behavior characteristics, users passing through, and users lingering. This is to account for situations where pedestrians may pass through the route but not for subway travel, or rest or hold activities near the subway station. Therefore, further filtering is performed based on mobile signaling data timestamp information. If a user is continuously captured by the same base station for more than 5 minutes, they are considered to be lingering or resting nearby and are not included in the passenger flow entering or exiting the station. For users captured near the station, if their information is captured by base stations corresponding to other travel routes, they are considered to be passing through the area and are also not included in the passenger flow entering or exiting the station. This is used to determine the activity behavior of people near the target station. User behavior characteristics are classified based on different time periods under different date types, such as... Figure 2 The image shows the number of people participating in various activities at different times during weekdays, such as... Figure 3 The figure shows the number of people participating in various activities at different times on non-working days, such as... Figure 4The chart shown is a line graph showing the number of people passing through and staying at different times on weekdays.

[0122] By classifying and statistically analyzing data according to different date types, the number of people engaged in various activities near the target station is obtained. The mobile signaling data generated by associated base stations within the target station's influence area is deduplicated, and the number of people engaged in various activities near the station is removed. This yields user data for those highly likely to take the subway near the station. Based on this user data, big data technology is used to predict and statistically analyze the signaling trajectory for the next 30 minutes, obtaining the travel trajectory information of users highly likely to take the subway within the corresponding time period. The target users taking the subway are identified, and it is determined whether the base station path of the target user deviates from the base station path of the target subway station within a preset time period. If the base station path of the target user deviates from the base station path of the target subway station, the target user who deviates from the base station path after the target subway station is included in the number of people exiting the target subway station.

[0123] Please refer to Figure 5 Specifically, methods for obtaining historical exit passenger flow data include:

[0124] Use the ARcGis tool to spatialize the geographic location of the target subway station;

[0125] The distance that can be reached within ten minutes on foot (approximately 800-1500m) is defined as the corresponding area of ​​influence;

[0126] The remaining information points within the influence range of each subway station, excluding the current subway station, are aggregated, and the base station information corresponding to the information points is obtained based on the location information of the base stations in the entire region, i.e., associated base stations;

[0127] The original signaling data is obtained from the associated base stations, and incomplete, abnormal, and invalid fields are removed.

[0128] The date features are divided according to the timestamp information in the target user's mobile phone signaling data to obtain the target user's mobile phone signaling data for different time periods. The date features include holidays, weekdays and weekends.

[0129] Behavioral feature analysis is performed on the timestamp information of the mobile phone signaling data of all users corresponding to the target subway station to obtain target users who meet the behavior characteristics of taking the subway, that is, target users who are highly likely to take rail transit, and other user data are eliminated; wherein the behavioral characteristics include: taking the subway, fast walking, slow walking, waiting, and lingering;

[0130] Based on the user's mobile phone signaling data, the base station path of the user is formed within a specific time period in the subway operation information. If the base station path of the user is the same as the base station path of the target subway station, it is considered that the target user traveled by subway to the target subway station, and the data of users who did not travel by subway to the target subway station are removed.

[0131] For the target user who is identified to travel by subway, if the base station path of the user leaves the base station path of the subway station within a certain period of time after a specific time, the user will be counted as the number of people leaving the station at the station corresponding to the base station that captured the user's signaling data before leaving the base station path of the subway station. That is, the target user who leaves the base station path after the target subway station will be counted as the number of people leaving the target subway station.

[0132] The number of people exiting each subway station is confirmed using the method described above, and the statistical number of people exiting is correlated with the corresponding time period to form a time series of the number of people exiting the target subway station.

[0133] In this embodiment of the invention, optionally, obtaining the time series of exiting passengers at the target subway station further includes:

[0134] Obtain the time series of exit passenger numbers at at least one adjacent station of the target subway station;

[0135] The correlation between the target subway station and the adjacent stations was determined using the Pearson correlation coefficient.

[0136] Obtain the time series of outbound passenger numbers from adjacent stations that meet the preset correlation threshold;

[0137] The time series of exits at the target subway station is weighted based on the time series of exits at adjacent stations that meet the preset correlation threshold, resulting in a weighted time series of exits at the target subway station. This weighted time series of exits at the target subway station is then used as the final time series of exits at the target subway station.

[0138] In this embodiment of the invention, when analyzing the passenger flow at each station, considering that there may be a certain relationship between the number of passengers at adjacent stations, the passenger flow information of each station can be collected by mobile phone signaling data and the passenger flow of two or four adjacent stations of the target station can be summarized according to the location information, and the correlation between adjacent stations can be calculated by Pearson correlation coefficient.

[0139] Taking the time series X of the number of people exiting the target subway station and the time series Y of the number of people exiting the adjacent previous station as an example, the correlation between the two variables is calculated using the Pearson correlation coefficient. Specifically:

[0140] The population Pearson correlation coefficient between two variables is defined as the quotient of the product of the covariance and the standard deviation of the two variables, usually denoted by ρ, and is defined as follows: ;

[0141] The estimated covariance and standard deviation can be used to obtain the Pearson correlation coefficient, denoted by r, and its expression is shown below: ;

[0142] After calculating the correlation coefficient between adjacent sites, it can be done through methods such as... Figure 6 The range of values ​​shown indicates the correlation strength of the variables. Based on the correlation strength between adjacent stations, for some adjacent stations with moderate or higher correlation strength, the time series of the number of passengers leaving the adjacent stations are weighted according to the relevant system and added to the time series of the target station to form a weighted time series of the target station. This allows for early warning to adjacent stations when the passenger flow data of a certain station is too large, so as to take preventive measures, minimize passenger congestion, and do a good job in traffic management.

[0143] Step 12: Encode the historical outbound passenger flow data to obtain the target encoded time series;

[0144] In this embodiment of the invention, optionally, the historical exit passenger flow data further includes: passenger flow data of adjacent stations of the target subway station, wherein the passenger flow data of the adjacent stations includes: a sequence of the number of people exiting or entering the station within a first historical time period;

[0145] The historical exit passenger flow data is encoded to obtain a time series of the encoded exit passenger numbers, including:

[0146] The time series of exits at the target subway station and the passenger flow data of the adjacent stations are encoded respectively;

[0147] The time series of exits at the coded target subway station is integrated with the passenger flow data of the adjacent stations to obtain the target coded time series.

[0148] In this embodiment of the invention, by encoding and integrating the historical exit passenger flow data, it is possible to better perceive the order and position of elements in the time series of exit passenger numbers at the target subway station, specifically including:

[0149] The location of each item in the time series data of the number of passengers exiting the station at time h is encoded using trigonometric function location coding. [11(t-h+1),…, 11(t)]

[0150] Position encoding function ( ): ; ;

[0151] Where n=10000 is the position in the sequence, d is the embedding vector dimension, and i is the column index mapped to;

[0152] Value encoder ( The method uses a fully connected (dense layer) to perform non-linear activation on each input element;

[0153] Time encoder ( Encode the time information by first converting it into a vector: slice the data every 15 minutes and represent it with a numerical vector, while adding holiday information, using 1 to represent a holiday and 0 to represent a non-holiday [year, month, day, hour, minute, whether it is a holiday];

[0154] The process for integrating the complete historical time series of outbound passenger numbers should be as follows: ;

[0155] The collected time series of passenger flow data from other connected historical transportation platforms (train stations or airports, etc.), i.e., passenger flow data from the adjacent stations, are integrated into... ;

[0156] The time series characteristics of the number of passengers exiting the target subway station are integrated with the time series of the number of passengers exiting other transportation platforms (adjacent stations) to form an integrated time series of the target station: .

[0157] In this embodiment of the invention, the historical exit passenger flow data is encoded and integrated through a feature embedding layer, enabling the model to not only learn the features of the target station independently, but also to capture the potential impact of other related stations on passenger flow. This allows the model to capture complex passenger flow trends and patterns, improving the accuracy of the model's prediction of passenger flow at subway stations.

[0158] Step 13: Obtain the dependencies between data in the target encoded time series, extract features from the data in the target encoded time series, and obtain the extracted features; based on the dependencies and the extracted features, predict the exit passenger flow data of the target subway station within a preset time period.

[0159] In this embodiment of the invention, the Transformer model is introduced, such as... Figure 7 As shown, the target encoded time series is input into an encoder layer, which consists of multiple sub-encoder layers. Each sub-encoder layer contains an attention mechanism and a feedforward neural network layer. The attention mechanism allows the encoder layer to interact and correlate elements in the input sequence, capturing the dependencies between elements. The feedforward neural network layer performs nonlinear transformation and feature extraction on the output of the encoder layer. The output of each sub-encoder layer is used as the input of the next sub-encoder layer, passing layer by layer. The last output of the sub-encoder layer is used as the input of the decoder layer. Correspondingly, the decoder layer also consists of multiple sub-decoder layers, used to gradually generate prediction results. Each time step of the sub-decoder layer can interact with the input sequence based on the prediction results of the previous time step and the attention mechanism. The last output of the sub-decoder layer is mapped to the dimension of the prediction result through a fully connected layer, and finally outputs the predicted value of the exit passenger flow data of the target subway station within a preset time period. By simultaneously focusing on global and local information in the input sequence, and integrating information from different locations through a multi-head attention mechanism, the relationship between elements in the sequence can be better captured, improving the accuracy of model prediction.

[0160] In this embodiment of the invention, historical exit passenger flow data is acquired, including: a time series of exit passenger numbers at a target subway station; the time series of exit passenger numbers at the target subway station is a sequence composed of exit passenger numbers at multiple preset time points within a first historical period; the historical exit passenger flow data is encoded to obtain a target encoded time series; the dependencies between data in the target encoded time series are obtained, and features are extracted from the data in the target encoded time series to obtain extracted features; based on the dependencies and extracted features, the predicted value of exit passenger flow data at the target subway station within a preset future time period is predicted. By simultaneously focusing on the global and local information in the time series of exit passenger numbers at the target subway station, information from different locations is integrated, thereby better capturing the relationships between elements in the sequence, improving prediction accuracy, and solving the problem of existing methods that cannot efficiently and accurately predict the number of subway passengers.

[0161] In this embodiment of the invention, optionally, a pre-trained subway station passenger flow prediction model is used to perform the following: encoding the historical exit passenger flow data to obtain a target encoded time series; obtaining the dependencies between data in the target encoded time series; extracting features from the data in the target encoded time series to obtain extracted features; and predicting the exit passenger flow data of the target subway station within a preset time period based on the dependencies and extracted features.

[0162] In this embodiment of the invention, optionally, the subway station passenger flow prediction model is trained based on a training set, which includes: historical exit passenger flow data samples and the actual values ​​of exit passenger flow data of the target subway station. The historical exit passenger flow data samples include a sequence of the number of people exiting the target subway station in a second historical period, and the actual values ​​of the exit passenger flow data of the target subway station are the exit passenger flow data of the target subway station in a third historical period, which is later than the second historical period.

[0163] In this embodiment of the invention, historical outbound passenger flow data is encoded by a model to ensure that the input data can be effectively processed and important time series features are preserved. Furthermore, by encoding the dependencies between data in the time series, the model identifies the mutual influence between different time points, enhancing the accuracy of prediction. Training based on historical outbound passenger flow data samples and real values ​​enables the model to learn effective patterns in a changing passenger flow environment, improving the accuracy of future passenger flow predictions. Moreover, through multi-level feature extraction, the model can learn complex and nonlinear relationships in the time series data, further enhancing the model's prediction accuracy.

[0164] In this embodiment of the invention, optionally, the subway station passenger flow prediction model includes: a feature embedding layer, an encoder layer, and a decoder layer;

[0165] The feature embedding layer encodes the historical outbound passenger flow data to obtain the target encoded time series;

[0166] The encoder layer obtains the dependencies between data in the target encoded time series, performs feature extraction on the data in the target encoded time series, and obtains the extracted features;

[0167] Based on the dependencies and extracted features, the decoder layer predicts the exit passenger flow data of the target subway station within a preset time period.

[0168] In this embodiment of the invention, optionally, the encoder layer includes multiple sub-encoder layers, the output of each sub-encoder layer is used as the input of the next sub-encoder layer, and the output of the last sub-encoder layer is used as the input of the decoder layer;

[0169] The decoder layer includes multiple sub-decoder layers. The output of each sub-decoder layer is used as the input of the next sub-decoder layer. The last sub-decoder layer outputs the predicted value of the exit passenger flow data of the target subway station within a preset time period.

[0170] Each sub-encoder layer includes an attention mechanism and a feedforward neural network layer; the attention mechanism interacts and correlates the time series data of the input to the sub-encoder layer to capture the dependencies between the data; the feedforward neural network layer extracts features from the output of the sub-encoder layer.

[0171] In this embodiment of the invention, the Transformer model is introduced, such as... Figure 7 As shown, the target encoded time series is input into an encoder layer, which consists of multiple sub-encoder layers. Each sub-encoder layer contains an attention mechanism and a feedforward neural network layer. The attention mechanism allows the encoder layer to interact and correlate elements in the input sequence, capturing the dependencies between elements. The feedforward neural network layer performs nonlinear transformation and feature extraction on the output of the encoder layer. The output of each sub-encoder layer is used as the input of the next sub-encoder layer, passing layer by layer. The last output of the sub-encoder layer is used as the input of the decoder layer. Correspondingly, the decoder layer also consists of multiple sub-decoder layers, used to gradually generate prediction results. Each time step of the sub-decoder layer can interact with the input sequence based on the prediction results of the previous time step and the attention mechanism. The last output of the sub-decoder layer is mapped to the dimension of the prediction result through a fully connected layer, and finally outputs the predicted value of the exit passenger flow data of the target subway station within a preset time period. By simultaneously focusing on global and local information in the input sequence, and integrating information from different locations through a multi-head attention mechanism, the relationship between elements in the sequence can be better captured, improving the accuracy of model prediction.

[0172] In this embodiment of the invention, optionally, it also includes:

[0173] The predicted passenger flow data of the target subway station within a preset time period is analyzed to generate a passenger flow planning, guidance and scheduling scheme.

[0174] In this embodiment of the invention, an intelligent system can analyze the predicted passenger flow data of the target subway station within a preset time period to guide and schedule the entrance and exit traffic of the target subway station. For example, during peak hours, more security checkpoints, verification channels, and platform space can be allocated based on the prediction results to cope with the increased traffic. The opening and closing status of entrances and exits can also be dynamically adjusted according to the actual situation to balance passenger flow. Digital technology can also be used to provide passengers with real-time passenger flow information, such as informing passengers of the congestion situation at different entrances and exits through in-station screens, guiding passengers to choose the more unobstructed entrances and exits to reduce crowd concentration. Furthermore, an intelligent queuing management system can be set up at the entrances and exits to adjust the size and location of queuing areas based on predicted traffic and actual conditions. This helps avoid excessively long queues, reduces passenger waiting time, and provides data support to relevant management and operation departments by comprehensively analyzing the influencing factors of subway passenger flow and improving the credibility and authority of the output results.

[0175] Please refer to Figure 8 This invention provides a subway station passenger flow prediction device, comprising:

[0176] The acquisition module 81 is used to acquire historical exit passenger flow data, which includes: a time series of the number of people exiting the target subway station; the time series of the number of people exiting the target subway station is a sequence composed of the number of people exiting the target subway station at multiple preset time points within a first historical period.

[0177] The first processing module 82 is used to encode the historical outbound passenger flow data to obtain the target encoded time series;

[0178] The second processing module 83 is used to obtain the dependency relationship between data in the target encoded time series, extract features from the data in the target encoded time series, and obtain the extracted features; based on the dependency relationship and the extracted features, predict the predicted value of the exit passenger flow data of the target subway station within a future preset time.

[0179] In this embodiment of the invention, optionally, the historical exit passenger flow data further includes: passenger flow data of adjacent stations of the target subway station, wherein the passenger flow data of the adjacent stations includes: a sequence of the number of people exiting or entering the station within a first historical time period;

[0180] The first processing module includes:

[0181] The first processing submodule is used to encode the time series of the number of people exiting the target subway station and the passenger flow data of the adjacent stations respectively; and to integrate the time series of the number of people exiting the target subway station and the passenger flow data of the adjacent stations to obtain the target encoded time series.

[0182] In this embodiment of the invention, optionally, the acquisition module includes:

[0183] The first acquisition submodule is used to acquire the target base station associated with the target subway station and acquire the mobile phone signaling data of the user under the target base station during the first historical time period;

[0184] The second processing submodule is used to obtain target users with subway riding behavior characteristics based on the mobile phone signaling data of users under the target base station in the first historical time period.

[0185] The third processing submodule is used to determine whether the base station that the target user passes through is a base station along the subway route within a preset time after the target user arrives at the target subway station. If the base station that the target user passes through is not a base station along the subway route, then the target user is counted as the number of people exiting the target subway station.

[0186] The fourth processing submodule is used to associate the number of people exiting the target subway station within the first historical period with a preset time point to form a time series of the number of people exiting the target subway station.

[0187] In this embodiment of the invention, optionally, the second processing submodule is configured to include:

[0188] The fifth processing submodule is used to acquire mobile phone signaling data of all users under the target base station within the first historical time period; perform behavioral feature analysis on the timestamp information in the mobile phone signaling data of all users to determine the type of the user, which includes: target users with subway riding behavior characteristics, users passing through, and users staying.

[0189] In this embodiment of the invention, optionally, the fifth processing submodule includes:

[0190] The sixth processing submodule is used to determine the user's travel trajectory based on the timestamp information in the mobile phone signaling data of all users; compare the user's travel trajectory with the subway transportation route; if the similarity between the user's travel trajectory and the subway transportation route meets a preset threshold, then the user is considered to be a target user with subway riding behavior characteristics.

[0191] In this embodiment of the invention, optionally, the historical outbound passenger flow data includes historical outbound passenger flow data with multiple different date characteristics, including holidays, weekdays, and weekends.

[0192] In this embodiment of the invention, optionally, the acquisition module further includes:

[0193] The second acquisition submodule is used to acquire the time series of the number of people exiting at at least one adjacent station of the target subway station;

[0194] The seventh processing submodule is used to determine the correlation between the target subway station and the adjacent stations using the Pearson correlation coefficient;

[0195] The third acquisition submodule is used to acquire the time series of the number of people leaving adjacent stations that meet the preset correlation threshold;

[0196] The eighth processing submodule is used to weight the time series of exits of the target subway station based on the time series of exits of adjacent stations that meet the preset correlation threshold, to obtain the weighted time series of exits of the target subway station, and to use the weighted time series of exits of the target subway station as the final determined time series of exits of the target subway station.

[0197] In this embodiment of the invention, optionally, a pre-trained subway station passenger flow prediction model is used to perform the following: encoding the historical exit passenger flow data to obtain a target encoded time series; obtaining the dependencies between data in the target encoded time series; extracting features from the data in the target encoded time series to obtain extracted features; and predicting the exit passenger flow data of the target subway station within a preset time period based on the dependencies and extracted features.

[0198] In this embodiment of the invention, optionally, the subway station passenger flow prediction model includes: a feature embedding layer, an encoder layer, and a decoder layer;

[0199] The feature embedding layer encodes the historical outbound passenger flow data to obtain the target encoded time series;

[0200] The encoder layer obtains the dependencies between data in the target encoded time series, performs feature extraction on the data in the target encoded time series, and obtains the extracted features;

[0201] Based on the dependencies and extracted features, the decoder layer predicts the exit passenger flow data of the target subway station within a preset time period.

[0202] In this embodiment of the invention, optionally, the encoder layer includes multiple sub-encoder layers, the output of each sub-encoder layer is used as the input of the next sub-encoder layer, and the output of the last sub-encoder layer is used as the input of the decoder layer;

[0203] The decoder layer includes multiple sub-decoder layers. The output of each sub-decoder layer is used as the input of the next sub-decoder layer. The last sub-decoder layer outputs the predicted value of the exit passenger flow data of the target subway station within a preset time period.

[0204] Each sub-encoder layer includes an attention mechanism and a feedforward neural network layer; the attention mechanism interacts and correlates the time series data of the input to the sub-encoder layer to capture the dependencies between the data; the feedforward neural network layer extracts features from the output of the sub-encoder layer.

[0205] In this embodiment of the invention, optionally, the subway station passenger flow prediction model is trained based on a training set, which includes: historical exit passenger flow data samples and the actual values ​​of exit passenger flow data of the target subway station. The historical exit passenger flow data samples include a sequence of the number of people exiting the target subway station in a second historical period, and the actual values ​​of the exit passenger flow data of the target subway station are the exit passenger flow data of the target subway station in a third historical period, which is later than the second historical period.

[0206] In this embodiment of the invention, optionally, it also includes:

[0207] The analysis module is used to analyze the predicted values ​​of passenger flow data of the target subway station within a preset time period, and generate passenger flow planning, guidance and scheduling schemes.

[0208] The subway station passenger flow prediction device provided in this embodiment of the invention can achieve Figure 1 The various processes implemented in the method embodiments achieve the same technical effect, and will not be described again here to avoid repetition.

[0209] This invention provides an electronic device 90, see [link to relevant documentation]. Figure 9 As shown, Figure 9 This is a schematic diagram of the electronic device 90 according to an embodiment of the present invention, including a processor 91, a memory 92, and a program or instructions stored in the memory 92 and executable on the processor 91. When the program or instructions are executed by the processor, they implement the steps in any of the subway station passenger flow prediction methods of the present invention.

[0210] This invention provides a readable storage medium on which a program or instruction is stored. When the program or instruction is executed by a processor, it implements the training method of any subway station passenger flow prediction model of this invention, or implements each process in any subway station passenger flow prediction method of this invention, and can achieve the same technical effect. To avoid repetition, it will not be described again here.

[0211] This application also provides a computer program product, including computer instructions, which, when executed by a processor, implement the above-described... Figure 1 The various processes of the method embodiments shown can achieve the same technical effect, and will not be described again here to avoid repetition.

[0212] Computer-readable media include both permanent and non-permanent, removable and non-removable media, which can store information using any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0213] It should be noted that the collection, gathering, updating, analysis, processing, use, transmission, and storage of user personal information involved in this disclosed technical solution all comply with relevant laws and regulations, are used for legitimate purposes, and do not violate public order and good morals. Necessary measures are taken to prevent unauthorized access to user personal information data and to safeguard user personal information security and network security.

[0214] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0215] The sequence numbers of the above embodiments of the present invention are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0216] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) and includes several instructions to cause a service classification device (which may be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in the various embodiments of this application.

[0217] The above description is only a preferred embodiment of this application. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of this application, and these improvements and modifications should also be considered within the scope of protection of this application.

Claims

1. A method for predicting passenger flow at subway stations, characterized in that, include: Obtain historical exit passenger flow data, which includes: a time series of the number of people exiting the target subway station; The time series of the number of people exiting the target subway station is a sequence composed of the number of people exiting the target subway station at multiple preset time points within a first historical period. The historical outbound passenger flow data is encoded to obtain the target encoded time series; The dependencies between data in the target encoded time series are obtained, and features are extracted from the data in the target encoded time series to obtain the extracted features. Based on the dependencies and the extracted features, the predicted value of the exit passenger flow data of the target subway station within a future preset time period is predicted.

2. The subway station passenger flow prediction method according to claim 1, characterized in that, The historical exit passenger flow data also includes: passenger flow data of adjacent stations of the target subway station, and the passenger flow data of the adjacent stations includes: a sequence of the number of people exiting or entering the station within the first historical period. The historical exit passenger flow data is encoded to obtain a time series of the encoded exit passenger numbers, including: The time series of exits at the target subway station and the passenger flow data of the adjacent stations are encoded respectively; The time series of exits at the coded target subway station is integrated with the passenger flow data of the adjacent stations to obtain the target coded time series.

3. The subway station passenger flow prediction method according to claim 1, characterized in that, Historical exit passenger flow data includes: Obtain the target base station associated with the target subway station, and obtain the mobile phone signaling data of the user under the target base station during the first historical time period; Based on the mobile phone signaling data of users under the target base station during the first historical period, target users with subway riding behavior characteristics are obtained. Determine whether the base station the target user passes through within a preset time after arriving at the target subway station is a base station along the subway route. If the base station the target user passes through is not a base station along the subway route, then the target user is counted as one of the exit passengers at the target subway station. The number of people exiting the target subway station during the first historical period is correlated with a preset time point to form a time series of the number of people exiting the target subway station.

4. The subway station passenger flow prediction method according to claim 3, characterized in that, The step of obtaining target users with subway-riding behavior characteristics based on mobile phone signaling data of users under the target base station during the first historical time period includes: Obtain mobile phone signaling data of all users under the target base station during the first historical time period; Behavioral feature analysis is performed on the timestamp information in the mobile phone signaling data of all users to determine the type of the user. The user types include: target users with subway riding behavior characteristics, users who pass through, and users who stay.

5. The subway station passenger flow prediction method according to claim 4, characterized in that, The step of performing behavioral feature analysis on the timestamp information in the mobile phone signaling data of all users to determine the type of the user includes: The user's travel trajectory is determined based on the timestamp information in the mobile phone signaling data of all users; The user's travel trajectory is compared with the subway route. If the similarity between the user's travel trajectory and the subway route meets a preset threshold, the user is considered a target user with subway riding behavior characteristics.

6. The subway station passenger flow prediction method according to claim 1, characterized in that, The historical exit passenger flow data includes historical exit passenger flow data with multiple different date characteristics, including holidays, weekdays, and weekends.

7. The subway station passenger flow prediction method according to claim 1, characterized in that, Obtaining the time series of exiting passengers at the target subway station also includes: Obtain the time series of exit passenger numbers at at least one adjacent station of the target subway station; The correlation between the target subway station and the adjacent stations was determined using the Pearson correlation coefficient. Obtain the time series of outbound passenger numbers from adjacent stations that meet the preset correlation threshold; The time series of exits at the target subway station is weighted based on the time series of exits at adjacent stations that meet the preset correlation threshold, resulting in a weighted time series of exits at the target subway station. This weighted time series of exits at the target subway station is then used as the final time series of exits at the target subway station.

8. The subway station passenger flow prediction method according to claim 1, characterized in that, Using a pre-trained subway station passenger flow prediction model, the following steps are performed: encoding the historical exit passenger flow data to obtain a target encoded time series; obtaining the dependencies between data in the target encoded time series; extracting features from the data in the target encoded time series to obtain extracted features; and predicting the exit passenger flow data of the target subway station within a preset time period based on the dependencies and extracted features.

9. The subway station passenger flow prediction method according to claim 8, characterized in that, The subway station passenger flow prediction model includes: a feature embedding layer, an encoder layer, and a decoder layer; The feature embedding layer encodes the historical outbound passenger flow data to obtain the target encoded time series; The encoder layer obtains the dependencies between data in the target encoded time series, performs feature extraction on the data in the target encoded time series, and obtains the extracted features; Based on the dependencies and extracted features, the decoder layer predicts the exit passenger flow data of the target subway station within a preset time period.

10. The subway station passenger flow prediction method according to claim 9, characterized in that, The encoder layer includes multiple sub-encoder layers, the output of each sub-encoder layer is used as the input of the next sub-encoder layer, and the output of the last sub-encoder layer is used as the input of the decoder layer. The decoder layer includes multiple sub-decoder layers. The output of each sub-decoder layer is used as the input of the next sub-decoder layer. The last sub-decoder layer outputs the predicted value of the exit passenger flow data of the target subway station within a preset time period. Each sub-encoder layer includes: an attention mechanism and a feedforward neural network layer; The attention mechanism interacts with and correlates the time-series data of the input to the sub-encoder layer, capturing the dependencies between the data; the feedforward neural network layer extracts features from the output of the sub-encoder layer.

11. The subway station passenger flow prediction method according to claim 8, characterized in that, The subway station passenger flow prediction model is trained based on a training set, which includes: historical exit passenger flow data samples and the actual values ​​of exit passenger flow data of the target subway station. The historical exit passenger flow data samples include a sequence of the number of people exiting the target subway station in a second historical period. The actual values ​​of the exit passenger flow data of the target subway station are the exit passenger flow data of the target subway station in a third historical period, which is later than the second historical period.

12. The subway station passenger flow prediction method according to claim 1, characterized in that, Also includes: The predicted passenger flow data of the target subway station within a preset time period is analyzed to generate a passenger flow planning, guidance and scheduling scheme.

13. A subway station passenger flow prediction device, characterized in that, include: The acquisition module is used to acquire historical exit passenger flow data, which includes: a time series of the number of people exiting the target subway station; the time series of the number of people exiting the target subway station is a sequence composed of the number of people exiting the target subway station at multiple preset time points within a first historical period; The first processing module is used to encode the historical outbound passenger flow data to obtain the target encoded time series; The second processing module is used to obtain the dependencies between data in the target encoded time series, extract features from the data in the target encoded time series, and obtain the extracted features; based on the dependencies and the extracted features, predict the predicted value of the exit passenger flow data of the target subway station within a future preset time.

14. An electronic device, characterized in that, It includes a processor, a memory, and a program or instructions stored in the memory and executable on the processor, wherein the program or instructions, when executed by the processor, implement the steps in the subway station passenger flow prediction method as described in any one of claims 1 to 12.

15. A readable storage medium, characterized in that, The readable storage medium stores a program or instructions that, when executed by a processor, implement the steps in the subway station passenger flow prediction method as described in any one of claims 1 to 12.

16. A computer program product, characterized in that, It includes computer instructions that, when executed by a processor, implement the steps in the subway station passenger flow prediction method as described in any one of claims 1 to 12.

Images