Method and device for determining cross-line dynamic ticket prices of rail transit passengers

Abstract

The invention discloses a rail transit passenger cross-line dynamic ticket price determination method and device, and the method comprises the steps: obtaining desensitized real-time operation data, sent by a data source layer, of a rail transit system; analyzing the real-time operation data to obtain predicted passenger flow volume data of the rail transit system in a preset time period; and the predicted passenger flow volume data are encrypted and transmitted to a central service layer, so that the central service layer adjusts the ticket price of the rail transit system in the preset time period according to the predicted passenger flow volume data. According to the invention, the cross-line dynamic ticket prices of the rail transit passengers can be adjusted through the central service layer and the federated node layer in a combined manner, so that the credibility and safety of the finally adjusted ticket prices in the preset time period can be ensured.

Description

TECHNICAL FIELD

[0001] Embodiments of the present application relate to the technical field of rail transit demand management, and in particular to a rail transit passenger cross-line dynamic fare determination method and device. BACKGROUND

[0002] In recent years, the rail transit network of major cities in China has continued to expand, and urban rail transit has become the backbone of many urban travel systems. With the year-on-year increase in passenger flow, the contradiction between passenger flow demand and supply capacity is becoming increasingly prominent, and the uneven spatio-temporal distribution of passenger flow exacerbates the supply shortage during peak hours, resulting in excessive load pressure on the line network, and the oversupply during off-peak hours leads to low utilization of line network capacity. In view of this phenomenon, time-of-day fares have become an effective means of managing peak-hour passenger flow demand in urban rail transit, and by differentiating the fares during peak and off-peak hours, the demand for peak-hour travel is stimulated to shift to off-peak hours.

[0003] However, when the traditional fare system calculates the time-of-day fare, it often needs to adjust the fare based on the predicted passenger flow data of the rail transit system; and when predicting the passenger flow data of the rail transit system, it generally needs to predict the passenger flow data based on data from multiple operators, such as real-time positioning data from subway operator A, ticketing databases from subway operator B, and transaction records from the UnionPay payment platform. Among them, when calculating data from different operators, the sensitive data from each operator cannot be directly shared, and when transmitting data from different operators, data tampering or leakage is likely to occur, which cannot guarantee that the fare of the rail transit system can be calculated based on data with high credibility and security. SUMMARY

[0004] The present application provides a rail transit passenger cross-line dynamic fare determination method and device, which can realize data security and credibility of rail transit passenger cross-line dynamic fare adjustment based on predicted passenger flow data calculated based on data from different operators.

[0005] In a first aspect, a rail transit passenger cross-line dynamic fare determination method is provided, applied to a federal node layer, the method comprising: obtaining real-time operation data of a rail transit system after desensitization processing sent by a data source layer; the real-time operation data includes service data provided by a first operator and real-time positioning data provided by a second operator; the service data includes stored ticketing record data and commuting regularity analysis data; the real-time positioning data includes real-time GPS coordinate data of rail transit vehicles and real-time passenger flow data of rail transit stations; analyzing the real-time operation data to obtain predicted passenger flow data of the rail transit system within a preset time period; The predicted passenger flow data is encrypted and transmitted to the central service layer, so that the central service layer adjusts the ticket price of the rail transit system in the preset time period according to the predicted passenger flow data.

[0006] Optionally, the joint board node layer is deployed with a first edge node and a second edge node; the real-time operation data is analyzed to obtain predicted passenger flow data of the rail transit system in a preset time period, including: According to the service data provided by the first operator and the first LSTM prediction model located in the first edge node, the passenger flow data of the rail transit system in the preset time period is predicted to obtain first predicted passenger flow sub-data; the first LSTM prediction model is obtained based on federated learning on the block chain by training a preset first LSTM model according to the ticket price time sequence characteristics corresponding to the historical operation data of the rail transit system and the historical passenger flow of the rail transit system in each time period; According to the real-time positioning data provided by the second operator and the second LSTM prediction model located in the second edge node, the passenger flow data of the rail transit system in the preset time period is predicted to obtain second predicted passenger flow sub-data; the second LSTM prediction model is obtained based on federated learning on the block chain by training a preset second LSTM model according to the passenger flow time sequence characteristics corresponding to the historical operation data of the rail transit system and the historical passenger flow of the rail transit system in each time period; According to the first predicted passenger flow sub-data and the second predicted passenger flow sub-data, the predicted passenger flow data is obtained.

[0007] Optionally, before the passenger flow data of the rail transit system in the preset time period is predicted according to the service data provided by the first operator and the first LSTM prediction model located in the first edge node to obtain the first predicted passenger flow sub-data, it further includes: According to the timestamp information, the service data provided by the first operator is aligned in time sequence in the first edge node, and the real-time positioning data provided by the second operator is aligned in time sequence in the second edge node to obtain aligned service data and real-time positioning data; According to the timestamp information, the service data provided by the first operator is aligned in time sequence in the first edge node, and the real-time positioning data provided by the second operator is aligned in time sequence in the second edge node to obtain aligned service data and real-time positioning data; In the first edge node, the aligned service data is extracted in time sequence to obtain the ticket price time sequence characteristics; The ticket price time sequence characteristics are input into the first LSTM prediction model to obtain the first predicted passenger flow sub-data of the rail transit system in the preset time period; Accordingly, according to the real-time positioning data provided by the second operator and the second LSTM prediction model located in the second edge node, the passenger flow data of the rail transit system in the preset time period is predicted to obtain second prediction passenger flow sub-data, including: In the second edge node, the aligned real-time positioning data is subjected to time sequence feature extraction to obtain passenger flow time sequence features; The passenger flow time sequence features are input into the second LSTM prediction model to obtain the second prediction passenger flow sub-data of the rail transit system in the preset time period.

[0008] Optionally, the real-time operation data further includes payment flow data provided by a UnionPay payment platform and municipal external data; before the ticket price time sequence features are input into the first LSTM prediction model to obtain the first prediction passenger flow sub-data of the rail transit system in the preset time period, further including: According to the timestamp information, the payment flow data and the municipal external data are subjected to data alignment processing in the first edge node in time sequence to obtain emergency event data aligned with the aligned service data; In the first edge node, the emergency event data is subjected to time sequence feature extraction to obtain municipal external features; Accordingly, the ticket price time sequence features are input into the first LSTM prediction model to obtain the first prediction passenger flow sub-data of the rail transit system in the preset time period, including: The ticket price time sequence features and the municipal external features are input into the first LSTM prediction model to obtain the first prediction passenger flow sub-data.

[0009] Optionally, the prediction passenger flow data is transmitted to the central service layer in an encrypted manner, so that the central service layer adjusts the ticket price of the rail transit system in the preset time period according to the prediction passenger flow data, including: The prediction passenger flow data is subjected to gradient clipping to obtain prediction passenger flow data subjected to gradient processing; The prediction passenger flow data subjected to gradient processing is injected with Gaussian noise to obtain prediction passenger flow data subjected to enhanced differential privacy; The prediction passenger flow data subjected to enhanced differential privacy is encrypted by using homomorphic encryption gradient to obtain encrypted prediction passenger flow data.

[0010] Optionally, further including: After the central service layer adjusts the ticket price in the preset time period, receiving passenger behavior feedback information sent by the application layer; the passenger behavior feedback information includes card swiping interval information, path deviation information and ticket price sensitivity information; According to the passenger behavior feedback information, the prediction passenger flow data of the rail transit system in the preset time period is adjusted to update the prediction passenger flow data.

[0011] In a second aspect, a method for determining a dynamic cross-line ticket price of a rail transit passenger is provided, and the method is applied to a central service layer, and the method comprises the following steps: obtaining predicted passenger flow data transmitted by a federal node layer in an encrypted manner; inputting the predicted passenger flow data into a ticket price optimization model located in the central service layer to obtain an adjustment result of a ticket price of each rail transit station in the rail transit system within a preset time period; the ticket price optimization model is obtained by training, based on federated learning, a local passenger flow and a basic ticket price of each historical time period on a block chain; sending the adjustment result of the ticket price of each rail transit station within the preset time period to an application layer, so that the application layer publishes the ticket price adjusted according to the adjustment result.

[0012] Optionally, the training method of the ticket price optimization model is as follows: obtaining model parameters of a local ticket price formulation model corresponding to each edge node in the federal node layer; the local ticket price formulation model is obtained by training, on an edge node, a local passenger flow and a basic ticket price corresponding to a rail transit station; aggregating the model parameters corresponding to each edge node by using a federated weighted average algorithm to generate the ticket price optimization model.

[0013] Optionally, the training method of the local ticket price formulation model is as follows: for each rail transit station, determining a time period type of the rail transit station in each historical time period according to a local passenger flow corresponding to the rail transit station in each historical time period; the time period type comprises a peak time period, a flat peak time period and an emergency mode; assigning an adjustment coefficient to a ticket price corresponding to the rail transit station in each historical time period according to the time period type of the rail transit station in each historical time period; in each edge node, training the adjustment coefficient and the basic ticket price corresponding to the ticket price of the rail transit station in each historical time period to obtain the local ticket price formulation model.

[0014] In a third aspect, a device for determining a dynamic cross-line ticket price of a rail transit passenger is provided, and the device is applied to a federal node layer, and the device comprises the following modules: a first data acquisition module, configured to acquire real-time operation data of a rail transit system transmitted by a data source layer after desensitization processing; the real-time operation data comprises service data provided by a first operator and real-time positioning data provided by a second operator; the service data comprises stored ticket record data and commuting habit analysis data; the real-time positioning data comprises real-time GPS coordinate data of a rail transit vehicle and real-time passenger flow data of a rail transit station; a prediction module configured to analyze the real-time operation data to obtain predicted passenger flow data of the rail transit system in a preset time period; an encrypted transmission module configured to encrypt and transmit the predicted passenger flow data to the central service layer, so that the central service layer adjusts the ticket price of the rail transit system in the preset time period according to the predicted passenger flow data.

[0015] In a fourth aspect, a device for determining a dynamic ticket price of a rail transit passenger across lines is provided, and the device is applied to a central service layer and includes: a second data acquisition module configured to acquire the predicted passenger flow data encrypted and transmitted by the federal node layer; a ticket price adjustment module configured to input the predicted passenger flow data encrypted and transmitted by the federal node layer into a ticket price optimization model located in the central service layer to obtain an adjustment result of the ticket price of each rail transit station in the rail transit system in the preset time period; the ticket price optimization model is obtained based on federated learning on a blockchain using local passenger flow and basic ticket price in each historical time period; an adjustment result sending module configured to send the adjustment result of the ticket price of each rail transit station in the preset time period to an application layer, so that the application layer publishes the ticket price adjusted according to the adjustment result.

[0016] In a fifth aspect, a federal node layer is provided, including a processor and a memory, the memory is configured to store a computer program, and the processor is configured to invoke and run the computer program stored in the memory to execute the method in the first aspect or the implementation manners thereof.

[0017] In a sixth aspect, a central service layer is provided, configured to store a computer program, and the computer program causes a computer to execute the method in the second aspect or the implementation manners thereof.

[0018] In a seventh aspect, a system for customizing a dynamic ticket price of a rail transit passenger across lines is provided, including a central service layer and a federal node layer.

[0019] Through the technical solutions provided in the present application, by analyzing the real-time operation data of the rail transit system sent by the data source layer after desensitization processing, the predicted passenger flow data of the rail transit system in a preset time period can be accurately determined based on the real-time operation data provided by different operators; the predicted passenger flow data is encrypted and transmitted to the central service layer to realize secure transmission of the predicted passenger flow data, so that the central service layer adjusts the dynamic ticket price of the rail transit passenger across lines according to the predicted passenger flow data in the preset time period, to ensure the credibility and security of the ticket price finally adjusted in the preset time period.

[0020] Other features and advantages of the present disclosure will be described in the following detailed description. BRIEF DESCRIPTION OF DRAWINGS

[0021] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following will briefly introduce the drawings needed to be used in the embodiment description. Obviously, the drawings in the following description are only some embodiments of the present application, and for those skilled in the art, other drawings can also be obtained without creative labor on the basis of these drawings.

[0022] Figure 1 An application scenario diagram provided by the embodiments of the present application; Figure 2 A flowchart of a determination method of a cross-line dynamic ticket price of a rail transit passenger provided by the embodiments of the present application; Figure 3 A flowchart of another determination method of a cross-line dynamic ticket price of a rail transit passenger provided by the embodiments of the present application; Figure 4 A schematic diagram of a determination device of a cross-line dynamic ticket price of a rail transit passenger provided by the embodiments of the present application; Figure 5 A schematic diagram of another determination device of a cross-line dynamic ticket price of a rail transit passenger provided by the embodiments of the present application; Figure 6 A schematic block diagram of a federation node layer provided by the embodiments of the present application. DETAILED DESCRIPTION

[0023] The technical solutions in the embodiments of the present application will be described clearly and completely in the following with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some of the embodiments of the present application, but not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0024] It should be noted that the terms “first”, “second”, and the like in the specification and claims of the present application and the above drawings are used to distinguish similar objects, and do not necessarily indicate a specific order or a chronological sequence. It should be understood that the data thus used can be interchanged under appropriate circumstances, so that the embodiments of the present application described herein can be implemented in an order other than that illustrated or described herein. In addition, the terms “include” and “have” and any variations thereof are intended to cover non-exclusive inclusion, for example, a process, method, system, product or server including a series of steps or units does not necessarily have to be limited to those steps or units clearly listed, but can include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0025] With the development of urban rail transit networks, cross-line passenger flow now accounts for over 35%. Traditional fare systems suffer from two major flaws: static pricing and data silos. Static pricing arises because fixed fare systems cannot adapt to dynamic passenger flow; data silos result from each operator setting its own fares, preventing the direct sharing of sensitive data. Therefore, how to securely and reliably determine fares within a rail transit system using data from different operators urgently requires further research.

[0026] To address the aforementioned technical problems, the inventive concept of this application is as follows: Real-time operational data of the rail transit system, after de-identification processing, is obtained from the data source layer; the real-time operational data includes service data provided by a first operator and real-time location data provided by a second operator; the service data includes stored ticketing record data and commuting pattern analysis data; the real-time location data includes real-time GPS coordinate data of rail transit vehicles and real-time passenger flow data of rail transit stations; the real-time operational data is analyzed to obtain predicted passenger flow data of the rail transit system within a preset time period; the predicted passenger flow data is encrypted and transmitted to the central service layer, enabling the central service layer to adjust the fare of the rail transit system within the preset time period based on the predicted passenger flow data. This application analyzes the anonymized real-time operation data of the rail transit system sent by the data source layer. Based on real-time operation data provided by different operators, it can accurately determine the predicted passenger flow data of the rail transit system within a preset time period. The predicted passenger flow data is then encrypted and transmitted to the central service layer to achieve secure transmission. This ensures the data security and reliability of the central service layer's dynamic fare adjustment for rail transit passengers across lines based on the predicted passenger flow data within the preset time period, thereby ensuring the reliability and security of the final adjusted fare within the preset time period.

[0027] It should be understood that the technical solution of this application can be applied to the following scenarios, but is not limited to: In some possible ways, Figure 1 An application scenario diagram provided for an embodiment of this application, such as Figure 1 As shown, this application scenario may include a federation node layer 110, a central service layer 120, and a data source layer 130. The federation node layer 110 can establish connections with the central service layer 120 and the data source layer 130 respectively via wired or wireless networks.

[0028] For example, the federated node layer 110 can be a desktop computer, laptop computer, tablet computer, etc., but is not limited to these. In one embodiment of this application, the federated node layer 110 can send a request message to the data source layer 130. This request message can be used to request the acquisition of anonymized real-time operation data of the rail transit system sent by the data source layer. Further, the federated node layer 110 can receive a response message sent by the data source layer 130. This response message includes the acquisition of the anonymized real-time operation data of the rail transit system sent by the data source layer. In another embodiment of the present invention, the central service layer 120 can send a request message to the federated node layer 110. This request message can be used to encrypt and transmit predicted passenger flow data to the central service layer 120, so that the central service layer 120 can adjust the fare of the rail transit system within a preset time period based on the predicted passenger flow data.

[0029] also, Figure 1 An exemplary embodiment is provided, consisting of a federated node layer 110, a central service layer 120, and a data source layer 130. In practice, other numbers of federated node layers, central service layers, and data source layers may be included, and this application does not limit this.

[0030] In other possible implementations, the technical solution of this application may also be executed by the aforementioned federal node layer 110, or by the aforementioned central service layer 120 or data source layer 130. This application does not impose any restrictions on this.

[0031] After introducing the application scenarios of the embodiments of this application, the technical solution of this application will be described in detail below: Figure 2 The flowchart illustrates a method for determining dynamic fares for rail transit passengers across lines, provided as an embodiment of this application. This method is applied to the federated node layer and can be used by, for example... Figure 1 The federated node layer 110 shown is executed, but is not limited to this. For example... Figure 2 As shown, the method may include the following steps: S210. Obtain the real-time operation data of the rail transit system sent by the data source layer after de-identification processing.

[0032] The real-time operational data includes service data provided by the first operator and real-time location data provided by the second operator; the service data includes stored ticketing record data and commuting pattern analysis data; the real-time location data includes real-time GPS coordinate data of rail transit vehicles and real-time passenger flow data of rail transit stations; the commuting pattern analysis data can be OD commuting pattern analysis data; and the stored ticketing record data can include cross-line transfer signs.

[0033] The OD commuting regularity analysis data includes basic OD data, time dimension data and population characteristic data, wherein the basic OD data includes geographic locations of origins and destinations, commuting time distribution (morning peak, evening peak, etc.), and commuting distance distribution; the time dimension data includes commuting mode comparison between weekdays and weekends, seasonal variation regularity, and special commuting mode during holidays; and the population characteristic data includes commuting characteristics of different age groups, association between occupation and commuting mode, and income level and traffic mode selection.

[0034] When the ticket record data is stored as cross-line transfer identification, the ticket record data includes basic transfer information, spatial location information and time-related information, wherein the basic transfer information includes transfer line identification, line color identification and transfer direction indication, the transfer line identification is the line name / number that can be transferred (such as "transfer line 2"), the line color identification is used for visual differentiation of standard color for each line, and the transfer direction indication is used to indicate the driving direction of each transfer line (such as "to XX direction"); the spatial location information includes transfer passage location, transfer distance data, transfer route map and floor distribution information, the transfer passage location is provided with an arrow indicating the direction and path of the transfer passage, and the transfer distance data is used to display the time or distance required for walking transfer (such as "walking about 3 minutes"); the transfer route map is a simplified transfer path schematic diagram, and the floor distribution information is used to indicate the platform layer where different lines are located (such as "B2 layer: line 1; B3 layer: line 2"); the time-related information includes first and last train time, train arrival information and peak period prompt, the first and last train time is the first and last train time of each transfer line at the station, the train arrival information is the real-time or predicted arrival time of the off-duty train (in the intelligent identification system), and the peak period prompt is the congestion period prompt of the transfer passage or line.

[0035] Here, the service data provided by the first operator can be used to analyze the travel regularity information of passengers; and the real-time positioning data provided by the second operator can be used to monitor the position of the rail transit vehicle and the passenger flow distribution in real time. The real-time passenger flow data of the rail transit station in the real-time positioning data can be understood as the real-time passenger flow data corresponding to each rail transit station in the rail transit system.

[0036] S220, analyzing the real-time operation data to obtain predicted passenger flow data of the rail transit system in a preset time period.

[0037] In this step, the real-time operation data can be analyzed to predict the passenger flow data of the rail transit system in the future preset time period, so as to improve the response efficiency of the prediction of the predicted passenger flow data, and at the same time, the predicted passenger flow data analyzed by the real-time operation data can be quickly adjusted to match the fare adjustment result of the sudden event.

[0038] In this step, the real-time operation data can be analyzed to predict the passenger flow data of the rail transit system in the future preset time period, so as to improve the response efficiency of the prediction of the predicted passenger flow data, and at the same time, the predicted passenger flow data analyzed by the real-time operation data can be quickly adjusted to match the fare adjustment result of the sudden event.

[0039] Here, "adjusting the fare in the preset time period" can be understood as adjusting the fare of the rail transit system in the preset time period based on the highest upper limit fare of the existing fare, for example, the highest upper limit fare of the rail transit system in city A is 7 yuan, and the upper limit of the fare adjustment is 7 yuan when the fare in the preset time period is adjusted.

[0040] In addition, in order to avoid the interests of residents living around the rail transit station and white-collar workers who cannot avoid peak travel being damaged after the fare adjustment, the real-time operation data can be analyzed by the federal node layer to identify the rigid travel group, such as fixed commuting residents, and the travel records of individuals in the group are encrypted and desensitized, and the following compensation schemes are provided for individuals in the group: providing subsidies (such as monthly commuting discount coupons) to individuals in the group during peak hours; accumulating the travel points of individuals in the group during off-peak hours to exchange cash discounts; setting a "resident travel cost cap value" to calculate the excess part according to the benchmark fare.

[0041] In this step, by encrypting and transmitting the predicted passenger flow data to the central service layer, the rail transit passenger cross-line dynamic fare adjustment of the central service layer and the federal node layer can be realized, so as to ensure the credibility and data security of the final adjusted fare in the preset time period.

[0042] By analyzing the real-time operation data of the rail transit system sent by the data source layer, the predicted passenger flow data of the rail transit system in the preset time period can be accurately judged based on the real-time operation data provided by different operators; the predicted passenger flow data is encrypted and transmitted to the central service layer to realize the secure transmission of the predicted passenger flow data, so that the central service layer adjusts the data security and credibility of the rail transit passenger cross-line dynamic fare based on the predicted passenger flow data in the preset time period, so as to ensure the credibility and security of the final adjusted fare in the preset time period.

[0043] In some possible implementation embodiments, the slab node layer is deployed with a first edge node and a second edge node; analyzing real-time operation data to obtain predicted passenger flow data of the rail transit system in a preset time period can include the following steps: S310, according to the service data provided by the first operator and the first LSTM prediction model located in the first edge node, predicting the passenger flow data of the rail transit system in a preset time period to obtain first predicted passenger flow sub-data.

[0044] The first LSTM prediction model is obtained by training a preset first LSTM model based on federated learning on a blockchain according to the ticket price time sequence features corresponding to the historical operation data of the rail transit system and the historical passenger flow of the rail transit system in each time period.

[0045] Here, the ticket price time sequence features corresponding to the historical operation data of the rail transit system can be understood as the ticket price time sequence features corresponding to each rail transit station in the rail transit system; the historical passenger flow of the rail transit system in each time period can be understood as the historical passenger flow corresponding to each rail transit station in the rail transit system in each time period.

[0046] S320, according to the real-time positioning data provided by the second operator and the second LSTM prediction model located in the second edge node, predicting the passenger flow data of the rail transit system in a preset time period to obtain second predicted passenger flow sub-data.

[0047] The second LSTM prediction model is obtained by training a preset second LSTM model based on federated learning on a blockchain according to the passenger flow time sequence features corresponding to the historical operation data of the rail transit system and the historical passenger flow of the rail transit system in each time period.

[0048] Here, the passenger flow time sequence features corresponding to the historical operation data of the rail transit system can be understood as the passenger flow time sequence features corresponding to each rail transit station in the rail transit system.

[0049] Here, the first LSTM prediction model can be locally trained in the first edge node, and the second LSTM prediction model can be locally trained in the second edge node, wherein the first LSTM prediction model and the second LSTM prediction model can both use Xavier to initialize the LSTM weight matrix involved in the training of the LSTM prediction model in the training process, so that the first LSTM prediction model and the second LSTM prediction model are both adapted to the non-linear activation function, and Kaiming normal distribution is used to initialize the fully connected layer coefficients involved in the training of the LSTM prediction model to solve the gradient problem of the first LSTM prediction model and the second LSTM prediction model in the activation process.

[0050] In addition, before the first edge node trains the first LSTM prediction model and the second edge node trains the second LSTM prediction model, initial parameters of the first LSTM prediction model of the first edge node and initial parameters of the second LSTM prediction model of the second edge node can be written into the genesis block after being hashed and confused, to ensure the initial parameters are tamper-proof and traceable.

[0051] In the training of the preset first LSTM model and / or the second LSTM model, a loss function constructed according to a prediction error and a regularization term is further combined to train the preset first LSTM model and / or the second LSTM model, to avoid the first LSTM prediction model and / or the second LSTM prediction model trained to predict the passenger flow fluctuates sharply; and a formula corresponding to the loss function constructed according to the prediction error and the ticket price fluctuation penalty term is as follows: , is the predicted passenger flow corresponding to the current time period predicted by the first LSTM prediction model or the second LSTM prediction model, is the actual passenger flow corresponding to the current time period, is the prediction error, is the regularization term.

[0052] S330, obtaining the predicted passenger flow data according to the first predicted passenger flow sub-data and the second predicted passenger flow sub-data.

[0053] Here, by deploying the first edge node and the second edge node on the federal node layer 110, the first edge node and the second edge node respectively analyze the service data provided by the first operator and the real-time positioning data provided by the second operator, so that the service data provided by the first operator and the real-time positioning data provided by the second operator are respectively analyzed in the first edge node and the second edge node, and the privacy protection of the service data provided by the first operator and the real-time positioning data provided by the second operator is realized.

[0054] By using the above method, the passenger flow data of the rail transit system in a preset time period is predicted according to the service data provided by the first operator and the first LSTM prediction model located in the first edge node, and according to the real-time positioning data provided by the second operator and the second LSTM prediction model located in the second edge node, the service data provided by the first operator and the real-time positioning data provided by the second operator are respectively analyzed in the first edge node and the second edge node, so that the privacy protection between the service data provided by the first operator and the real-time positioning data provided by the second operator is realized.

[0055] In some possible implementation examples, before predicting the passenger flow data of the rail transit system in the preset time period according to the service data provided by the first operator and the first LSTM prediction model located in the first edge node to obtain the first predicted passenger flow sub-data, the method can further include: performing data alignment processing on the service data provided by the first operator in the first edge node and the real-time positioning data provided by the second operator in the second edge node in time sequence according to timestamp information to obtain the aligned service data and the real-time positioning data.

[0056] Here, the data alignment processing on the service data provided by the first operator and the real-time positioning data provided by the second operator can realize alignment and fusion of the data provided by the first operator and the second operator.

[0057] In the data alignment processing on the above data, the data alignment processing on the service data provided by the first operator and the real-time positioning data provided by the second operator can be performed through time window slicing, for example, 5-minute granularity, and at the same time, the privacy protection in the data alignment processing on the above data can be realized by combining a fuzzy hash technology.

[0058] Correspondingly, predicting the passenger flow data of the rail transit system in the preset time period according to the service data provided by the first operator and the first LSTM prediction model located in the first edge node to obtain the first predicted passenger flow sub-data can include the following steps. S410, performing time sequence feature extraction on the aligned service data in the first edge node to obtain a ticket price time sequence feature.

[0059] It should be noted that the ticket price time sequence feature can be a historical feature, for example, the ticket price time sequence feature can be a basic ticket price in the last time period; or the ticket price time sequence feature can be a current feature, for example, the ticket price time sequence feature can be a basic ticket price in the current time period.

[0060] The service data is subjected to basic timing feature extraction, that is, the aligned service data is subjected to time dimension decomposition and statistical features in turn. When the service data is subjected to time dimension decomposition, the time granularity aggregation mode can be used, that is, the average ticket price and the ticket price median are calculated according to hours / days / weeks / months, and then the service data is subjected to periodic decomposition to decompose the ticket price into trend, seasonality and residual components. Then, the sliding window statistics are performed on the service data to calculate the rolling average of the ticket price and the rolling standard deviation of the ticket price. Then, based on the average ticket price, the ticket price median, the trend, the seasonality, the residual component, the rolling average of the ticket price and the rolling standard deviation of the ticket price statistical features obtained above, specifically, the mean, the variance, the maximum value, the minimum value and the quantile of the ticket price are obtained by descriptive statistics, the percentage and the same period change rate of the ticket price are obtained based on the change feature statistics, and the distribution histogram, the skewness and the kurtosis of the ticket price are obtained by the distribution feature statistics. Finally, based on the mean, the variance, the maximum value, the minimum value and the quantile of the ticket price, the percentage and the same period change rate of the ticket price, and the percentage and the same period change rate of the ticket price, the basic ticket price in the last time period or the basic ticket price in the current time period is calculated, and the basic ticket price in the last time period or the basic ticket price in the current time period is taken as the ticket price timing feature.

[0061] S420, inputting the ticket price timing feature into the first LSTM prediction model to obtain the first predicted passenger flow sub-data of the rail transit system in the preset time period.

[0062] The LSTM prediction model is obtained based on the time series output features corresponding to the historical operation data of the rail transit system and the historical passenger flow of the rail transit system in each time period, and is trained on a blockchain based on federated learning.

[0063] By the above method, the ticket price timing feature is obtained by performing timing feature extraction on the aligned service data in the first edge node. The ticket price timing feature is input into the first LSTM prediction model located in the first edge node, so that the first predicted passenger flow sub-data of the rail transit system in the preset time period can be obtained in the first edge node, so as to realize the prediction of the passenger flow of the rail transit system in the preset time period in the first edge node.

[0064] Correspondingly, the passenger flow data of the rail transit system in the preset time period is predicted based on the real-time positioning data provided by the second operator and the second LSTM prediction model located in the second edge node to obtain the second predicted passenger flow sub-data, which can include the following steps: S430, in the second edge node, performing timing feature extraction on the aligned real-time positioning data to obtain passenger flow timing features.

[0065] Here, the passenger flow time sequence feature can be a historical feature, for example: the passenger flow time sequence feature is the passenger flow in the last time period; the passenger flow time sequence feature can also be the passenger flow in the current time period.

[0066] The indirect passenger flow feature extraction based on the aligned real-time positioning data and vehicle GPS data is performed on the aligned real-time positioning data, that is, the vehicle operation feature extraction and passenger flow inference are sequentially performed on the aligned real-time positioning data. In the vehicle operation feature extraction on the aligned real-time positioning data, the arrival interval feature of the aligned real-time positioning data can be extracted based on the interval time of consecutive vehicles arriving at the same station; the running speed feature of the aligned real-time positioning data can be extracted based on the average speed of the vehicle running between stations calculated by segments; and the stop time feature of the aligned real-time positioning data can be extracted based on the length of time the vehicle stays at the platform. Then, based on the arrival interval feature, the running speed feature and the stop time feature, the passenger flow inference is performed. Specifically, based on the arrival interval feature, the running speed feature and the stop time feature, the vehicle full load rate is calculated in combination with the vehicle type and the GPS speed change to obtain the estimated passenger load; based on the arrival interval feature, the running speed feature and the stop time feature, the passenger flow direction between stations is inferred according to the GPS data of multiple train services to calculate the passenger flow corresponding to the origin and destination; based on the arrival interval feature, the running speed feature and the stop time feature, the passenger flow aggregation point is identified through vehicle abnormal stop to detect abnormal aggregation to calculate the passenger flow of the aggregation point; finally, based on the estimated passenger load, the passenger flow corresponding to the origin and destination, and the passenger flow corresponding to the origin and destination, the passenger flow in the last time period or the passenger flow in the current time period is calculated, and the passenger flow in the last time period or the passenger flow in the current time period is taken as the passenger flow time sequence feature.

[0067] S440, inputting the passenger flow time sequence feature into the second LSTM prediction model to obtain the second predicted passenger flow sub-data of the rail transit system in the preset time period.

[0068] By using the above method, the passenger flow time sequence feature is obtained by performing time sequence feature extraction on the aligned real-time positioning data in the second edge node; by inputting the passenger flow time sequence feature into the second LSTM prediction model located in the second edge node, the second predicted passenger flow sub-data of the rail transit system in the preset time period can be obtained in the second edge node, so as to realize the prediction of the passenger flow of the rail transit system in the preset time period in the second edge node, and further make the aligned service data and real-time positioning data invisible to each other in the analysis process in the first edge node and the second edge node.

[0069] Further, the real-time operation data further includes payment flow data provided by the UnionPay payment platform and municipal external data.

[0070] Here, the municipal external data includes weather level data (such as heavy rain, high temperature) and abnormal event data (such as accidents, flow restrictions), which can be used to evaluate the influence of external factors on passenger flow; the payment stream data can be payment stream data after desensitization processing, and the payment stream data can be used to construct a passenger ticket price sensitive portrait to analyze the reaction of different passenger groups to ticket prices; the payment stream data is desensitized by the hash confusion technology to ensure that the payment stream data after desensitization meets the requirements of the Personal Information Security Specification.

[0071] Correspondingly, before inputting the ticket price time sequence feature into the first LSTM prediction model to obtain the first predicted passenger flow sub-data of the rail transit system in the preset time period, the following steps can also be included: S510, according to the timestamp information, the payment stream data and the municipal external data are aligned in the first edge node in time sequence to obtain the emergency event data aligned with the aligned service data.

[0072] Here, the payment stream data and the municipal external data are aligned to obtain the emergency event data aligned with the aligned service data, which can realize the alignment and fusion of the payment stream data, the municipal external data and the aligned service data. The emergency event data includes payment stream data and municipal external data aligned with the aligned service data.

[0073] Among them, when the above data is aligned, the payment stream data and the municipal external data can also be aligned in time sequence by the time window slicing as described above, and at the same time, the privacy protection in the data alignment process of the above data can also be realized by combining the fuzzy hash technology.

[0074] S520, in the first edge node, the time sequence feature of the emergency event data is extracted to obtain the municipal external feature.

[0075] Here, the municipal external feature can be a historical feature, for example: the municipal external feature is the weather condition in the last time period; the municipal external feature can also be the weather condition in the current time period.

[0076] Correspondingly, the ticket price time sequence feature is input into the first LSTM prediction model to obtain the first predicted passenger flow sub-data of the rail transit system in the preset time period, which also includes: inputting the ticket price time sequence feature and the municipal external feature into the first LSTM prediction model to obtain the first predicted passenger flow sub-data.

[0077] By the above method, in the first edge node, by inputting the municipal external features obtained by performing temporal feature extraction on the burst event data aligned with the aligned service data and the fare temporal features into the first LSTM prediction model, the first predicted passenger flow sub-data matched with the burst event can be predicted, so as to quickly adjust the fare adjustment result matched with the burst event, and thus the finally adjusted fare can cope with the occurrence of the burst event.

[0078] In some possible implementation embodiments, the encrypted prediction passenger flow data is transmitted to the central service layer, so that the central service layer adjusts the fare of the rail transit system in the preset time period according to the prediction passenger flow data, which can include the following steps: S610, gradient clipping is performed on the prediction passenger flow data to obtain prediction passenger flow data processed by gradient.

[0079] Here, the gradient threshold δ of 0.01 can be used to perform gradient clipping on the prediction passenger flow data to prevent data explosion of the prediction passenger flow data processed by gradient.

[0080] S620, the prediction passenger flow data processed by gradient is injected with Gaussian noise to obtain prediction passenger flow data with enhanced differential privacy.

[0081] Here, the noise variance of 0.1 can be used to inject Gaussian noise into the prediction passenger flow data processed by gradient, so that the prediction passenger flow data can be constrained by the privacy budget, and thus the differential privacy of the prediction passenger flow data with enhanced differential privacy is enhanced.

[0082] S630, the prediction passenger flow data with enhanced differential privacy is encrypted by using a homomorphic encryption gradient to obtain encrypted prediction passenger flow data.

[0083] Here, the prediction passenger flow data with enhanced differential privacy can be encrypted by using a Paillier homomorphic encryption gradient to ensure the transmission security of the prediction passenger flow data.

[0084] By the above method, by sequentially performing gradient clipping processing, Gaussian noise injection processing, and encryption processing by using a homomorphic encryption gradient on the prediction passenger flow data, the transmission security of the processed prediction passenger flow data can be improved.

[0085] In some possible implementation embodiments, the rail transit passenger cross-line dynamic fare determination method further includes: S710, after the central service layer adjusts the fare in the preset time period, receiving passenger behavior feedback information sent by the application layer.

[0086] The passenger behavior feedback information includes card swiping interval information, path deviation information and ticket price sensitivity information. The card swiping interval information includes time or space interval data of two consecutive card swiping of the passenger in the rail transit system, and is mainly used for analyzing passenger travel habits and behavior; the path deviation information is difference information between the actual travel path of the passenger and the shortest / conventional path recommended by the system; the ticket price sensitivity information is the sensitivity degree of the passenger travel selection to the ticket price change, and is used for evaluating the influence of the pricing strategy. The ticket price sensitivity information includes historical ticket price payment data and alternative selection, wherein the historical ticket price payment data is the travel frequency of the passenger under different ticket prices, and the alternative selection is whether to switch to other transportation tools (such as bus to shared bicycle) after the ticket price adjustment.

[0087] In S720, the predicted passenger flow data of the rail transit system in the preset time period is adjusted according to the passenger behavior feedback information, so as to update the predicted passenger flow data.

[0088] Here, the passenger behavior feedback information is feature extracted to realize the conversion of the passenger behavior feedback information into quantifiable indexes and the docking with the predicted passenger flow model. The card swiping interval information is feature extracted to obtain transfer features of the passenger, and the transfer features include transfer proportion, average transfer time, and abnormal card swiping frequency (such as repeated card swiping in a short time). Here, if the average transfer time is significantly increased, it may indicate that the passenger changes the path due to congestion or delay. According to the path deviation information, the deviation rate feature (such as the number of stations / time) of the actual path and the theoretical path is extracted to analyze the reasons (such as construction, temporary station closure). According to the ticket price sensitivity information, the passenger flow change rate feature (such as the increase of off-peak passenger flow when the discount is implemented) after the ticket price adjustment is extracted. Then, the predicted passenger flow data of the rail transit system in the preset time period is adjusted according to the correlation between the transfer features, the deviation rate features and the passenger flow change rate features and the predicted passenger flow in the predicted passenger flow model. For example, if the average transfer time in the transfer features is prolonged, it may be caused by vehicle congestion, and then the predicted passenger flow of the next three stations is increased by 5%-10% after adjustment. For another example, after the ticket price adjustment, if the passenger flow of the sensitive group (such as the elderly) decreases by 15% on weekends, the predicted value corresponding to the passenger flow on weekends is lowered.

[0089] By adjusting the predicted passenger flow data of the rail transit system in the preset time period according to the passenger behavior feedback information, the closed loop of the adjusted ticket price and the passenger behavior feedback can be established, and then the predicted passenger flow data can be adjusted through the behavior feedback of the passenger to the adjusted ticket price, so as to improve the accuracy of the final predicted passenger flow data.

[0090] Figure 3 A flowchart of a rail transit passenger cross-line dynamic ticket price determination method provided by the embodiment of the present application is shown in FIG. 1. The method is applied to a central service layer and can be implemented by a server such as a cloud server.Figure 1 The central service layer 120 shown performs, but is not limited to. As shown, the method can include the following steps: Figure 3 S810, obtaining the predicted passenger flow data encrypted and transmitted by the federal node layer.

[0091] S820, inputting the predicted passenger flow data into a fare optimization model located in the central service layer to obtain an adjustment result of the fare of each rail transit station in the rail transit system within a preset time period.

[0092] The fare optimization model is obtained by training the local passenger flow and the basic fare of each historical time period on the blockchain based on federated learning.

[0093] In addition, when training the fare optimization model, the fare optimization model is optimized based on a preset constraint condition to update the fare optimization model; wherein the preset constraint condition can be that the adjusted fare within the preset time period is 80% to 100% of the reference fare.

[0094] S830, sending the adjustment result of the fare of each rail transit station within the preset time period to the application layer, so that the application layer publishes the fare adjusted according to the adjustment result.

[0095] Here, the application layer can be deployed on a fare publishing terminal, such as a vending machine and a mobile phone APP, so that passengers can purchase according to the adjusted fare published by the fare publishing terminal.

[0096] With the above method, after inputting the predicted passenger flow data encrypted and transmitted by the federal node layer into the fare optimization model to obtain the adjustment result of the fare of each rail transit station in the rail transit system within the preset time period, the adjustment result of the fare of each rail transit station within the preset time period is sent to the application layer, so that the application layer publishes the fare adjusted according to the adjustment result. The final published adjusted fare within the preset time period takes into account the balance of passenger flow and the stability of fare.

[0097] In some possible implementation embodiments, the training method of the fare optimization model is as follows: S910, obtaining the model parameters of the local fare setting model corresponding to each edge node in the federal node layer.

[0098] The local fare setting model is obtained by training the local passenger flow and the basic fare of each historical time period corresponding to the rail transit station on the edge node.

[0099] ​In the process of obtaining the model parameters of the local ticket pricing model corresponding to each edge node in the federal node layer, the model parameters of the local ticket pricing model corresponding to each edge node can be obtained based on the smart contract of the Hyperledger Fabric chain deployed in the federal node layer after verifying the identity certificate of the participant, so as to prevent Sybil attacks in the network.

[0100] S920, using a federated weighted average algorithm to aggregate the model parameters corresponding to each edge node to generate a ticket pricing optimization model.

[0101] Here, the formula used to aggregate the model parameters corresponding to each edge node using the federated weighted average algorithm is as follows:

[0102] is the model parameter corresponding to the ticket pricing optimization model; is the model parameter corresponding to the edge node; is a weight dynamically calculated according to the proportion of the data amount corresponding to the service data provided by the first operator and the real-time positioning data provided by the second operator, for example, when the proportion of the data amount corresponding to the service data provided by the first operator is 40% and the proportion of the data amount corresponding to the service data provided by the second operator is 60%, the weight corresponding to the service data provided by the first operator is = 0.4, and the weight corresponding to the real-time positioning data provided by the second operator is = 0.6.

[0103] It should be noted that when the model parameter corresponding to the ticket pricing optimization model is stored on the chain, the model parameter corresponding to the ticket pricing optimization model can be hashed to realize SHA-256 encryption of the model parameter, so as to ensure the audit traceability of the model parameter.

[0104] By using the above method, the model parameters of the local ticket pricing model corresponding to each edge node are aggregated using the federated weighted average algorithm, which can realize the combination of federated learning and blockchain to generate the final ticket pricing optimization model, and further ensure the credibility and security of the ticket pricing optimization model training.

[0105] Further, the training method of the local ticket pricing model is as follows: S1010, for each rail transit station, according to the local passenger flow of the rail transit station in each historical time period, determine the time period type corresponding to the rail transit station in each historical time period.

[0106] Among them, the time period type includes peak period, flat peak period and emergency mode.

[0107] ​In this step, the local passenger flow of each rail transit station in each historical time period can determine the time period type corresponding to each time period, which provides the basis for adjusting the coefficient of ticket price distribution in step S1020 for each time period.

[0108] S1020, according to the time period type corresponding to the rail transit station in each historical time period, the ticket price corresponding to the rail transit station in each historical time period is distributed with an adjustment coefficient.

[0109] For example, by the local passenger flow of the rail transit station in historical time period A, historical time period B and historical time period C, it is determined that the rail transit station in historical time period A is peak time period, historical time period B is flat peak time period, and historical time period C is emergency mode. The ticket price adjustment coefficient corresponding to the rail transit station in historical time period A is 1, that is, the ticket price of the rail transit station in the time period corresponding to historical time period A is the basic ticket price; the ticket price adjustment coefficient corresponding to the rail transit station in historical time period B is 0.9, and the ticket price of the rail transit station in the time period corresponding to historical time period B is adjusted to 0.9 times of the basic ticket price; the ticket price adjustment coefficient corresponding to the rail transit station in historical time period C is 0.7~1, wherein the coefficient corresponding to the rail transit station in this time period also needs to be dynamically adjusted according to the event level.

[0110] S1030, in each edge node, the adjustment coefficient and the basic ticket price corresponding to the ticket price of the rail transit station in each historical time period are trained to obtain a local ticket price making model.

[0111] By using the above method, the local ticket price making model obtained can correspondingly make the local ticket price of the rail transit station in different time period types, so that the local ticket price obtained by the local ticket price making model is accurately customized under different local passenger flow of the rail transit station.

[0112] Figure 4 is a schematic diagram of a rail transit passenger cross-line dynamic ticket price determination device 1100 of an embodiment provided by the present application. The device 1100 is applied to a federal node layer, as shown in the figure, the device 800 includes: Figure 4 The first data acquisition module 1110 is configured to acquire the real-time operation data of the rail transit system sent by the data source layer after desensitization processing; the real-time operation data includes service data provided by a first operator and real-time positioning data provided by a second operator; the service data includes stored ticketing record data and commuting habit analysis data; the real-time positioning data includes real-time GPS coordinate data of rail transit vehicles and real-time passenger flow data of rail transit stations. ​The prediction module 1120 is configured to analyze the real-time operation data to obtain predicted passenger flow data of the rail transit system in a preset time period. The encrypted transmission module 1130 is configured to transmit the predicted passenger flow data to the central service layer in an encrypted manner, so that the central service layer adjusts the ticket price of the rail transit system in the preset time period according to the predicted passenger flow data.

[0113] In some implementations, the prediction module 1120 includes: The first predicted passenger flow sub-data obtaining unit is configured to predict the passenger flow data of the rail transit system in the preset time period according to the service data provided by the first operator and the first LSTM prediction model located in the first edge node, to obtain first predicted passenger flow sub-data; the first LSTM prediction model is obtained by training a preset first LSTM model based on federated learning on a blockchain according to the ticket price time sequence feature corresponding to the historical operation data of the rail transit system and the historical passenger flow of the rail transit system in each time period. The second predicted passenger flow sub-data obtaining unit is configured to predict the passenger flow data of the rail transit system in the preset time period according to the real-time positioning data provided by the second operator and the second LSTM prediction model located in the second edge node, to obtain second predicted passenger flow sub-data; the second LSTM prediction model is obtained by training a preset second LSTM model based on federated learning on a blockchain according to the passenger flow time sequence feature corresponding to the historical operation data of the rail transit system and the historical passenger flow of the rail transit system in each time period. The predicted passenger flow data obtaining unit is configured to obtain the predicted passenger flow data according to the first predicted passenger flow sub-data and the second predicted passenger flow sub-data.

[0114] In some implementations, the prediction module 1120 further includes: The first data alignment unit is configured to perform data alignment processing on the service data provided by the first operator in the first edge node and the real-time positioning data provided by the second operator in the second edge node in a time sequence according to the timestamp information, to obtain aligned service data and real-time positioning data. Correspondingly, the first predicted passenger flow sub-data obtaining unit includes: The first time sequence feature extraction sub-unit is configured to perform time sequence feature extraction on the aligned service data in the first edge node, to obtain a ticket price time sequence feature. The first predicted sub-data obtaining sub-unit is configured to input the ticket price time sequence feature into the first LSTM prediction model, to obtain the first predicted passenger flow sub-data of the rail transit system in the preset time period. Correspondingly, the second predicted passenger flow sub-data obtaining unit includes: The second time sequence feature extraction subunit is configured to perform time sequence feature extraction on the aligned real-time positioning data in the second edge node to obtain time sequence features of passenger flow; The second predicted sub-data obtaining subunit is configured to input the time sequence features of passenger flow into the second LSTM prediction model to obtain second predicted passenger flow sub-data of the rail transit system within a preset time period.

[0115] In some implementable manners, the real-time operation data further includes payment flow data provided by a UnionPay payment platform and municipal external data; and the prediction module 1120 further includes: The second data alignment unit is configured to perform data alignment processing on the payment flow data and the municipal external data in the first edge node according to the timestamp information to obtain emergency event data aligned with the aligned service data; The municipal external feature extraction unit is configured to perform time sequence feature extraction on the emergency event data in the first edge node to obtain municipal external features; Correspondingly, the first predicted sub-data obtaining subunit includes: The first predicted sub-data adjusting subunit is configured to input the ticket price time sequence features and the municipal external features into the first LSTM prediction model to obtain first predicted passenger flow sub-data.

[0116] In some implementable manners, the encrypted transmission module 1130 includes: The gradient processing unit is configured to perform gradient clipping on the predicted passenger flow data to obtain predicted passenger flow data after gradient processing; The differential privacy unit is configured to inject Gaussian noise into the predicted passenger flow data after gradient processing to obtain predicted passenger flow data with enhanced differential privacy; The encryption unit is configured to encrypt the predicted passenger flow data with enhanced differential privacy by using homomorphic encryption gradient to obtain predicted passenger flow data after encryption.

[0117] In some implementable manners, the device 1100 further includes: The feedback information receiving module is configured to receive passenger behavior feedback information sent by the application layer after the ticket price within the preset time period is adjusted in the central service layer; the passenger behavior feedback information includes card swiping interval information, path deviation information and ticket price sensitivity information; The data adjusting module is configured to adjust the predicted passenger flow data of the rail transit system within the preset time period according to the passenger behavior feedback information to update the predicted passenger flow data.

[0118] Figure 5 FIG. 12 is a schematic diagram of a determination device 1200 of a rail transit passenger cross-line dynamic ticket price according to an embodiment of the present application. The device 1200 is applied to a central service layer, such asFigure 5 As shown, the apparatus 900 includes: The second data acquisition module 1210 is configured to acquire the predicted passenger flow data transmitted by the federal node layer in an encrypted manner. The ticket price adjustment module 1220 is configured to input the predicted passenger flow data into a ticket price optimization model located in the central service layer to obtain an adjustment result of the ticket price of each rail transit station in the rail transit system within a preset time period. The ticket price optimization model is obtained by training the local passenger flow and the basic ticket price of each historical time period based on federated learning on the blockchain. The adjustment result sending module 1230 is configured to send the adjustment result of the ticket price of each rail transit station within the preset time period to the application layer, so that the application layer publishes the ticket price adjusted according to the adjustment result.

[0119] In some implementable manners, the apparatus 1200 further includes: The model parameter acquisition module is configured to acquire model parameters of a local ticket price formulation model corresponding to each rail transit station in the federal node layer. The local ticket price formulation model is obtained by training the local passenger flow and the basic ticket price of each historical time period corresponding to the rail transit station on the edge node. The ticket price optimization model generation module is configured to aggregate the model parameters corresponding to each edge node by using a federated weighted average algorithm to generate a ticket price optimization model.

[0120] In some implementable manners, the apparatus 1200 further includes: The time period type determination module is configured to determine, for each rail transit station, a time period type corresponding to the rail transit station in each historical time period according to the local passenger flow corresponding to the rail transit station in each historical time period. The time period type includes a peak time period, a flat peak time period, and an emergency mode. The adjustment coefficient determination module is configured to assign an adjustment coefficient to the ticket price corresponding to the rail transit station in each historical time period according to the time period type corresponding to the rail transit station in each historical time period. The local ticket price formulation model training module is configured to train, in each edge node, the adjustment coefficient and the basic ticket price corresponding to the rail transit station in each historical time period to obtain a local ticket price formulation model.

[0121] It should be understood that the rail transit passenger cross-line dynamic ticket price determination apparatus embodiment and the rail transit passenger cross-line dynamic ticket price determination method embodiment can correspond to each other, and similar descriptions can be referred to the rail transit passenger cross-line dynamic ticket price determination method embodiment. To avoid repetition, details are not described here. Specifically, Figure 4 and Figure 5The apparatus 1100 and the apparatus 1200 shown can perform the above-mentioned method embodiments of determining dynamic fares for rail transit passengers crossing lines, and the aforementioned and other operations and / or functions of the respective modules of the apparatus 1100 and the apparatus 1200 are respectively for realizing the corresponding procedures in the above-mentioned method of determining dynamic fares for rail transit passengers crossing lines, and for brevity, will not be repeated here.

[0122] The apparatus 1100 and the apparatus 1200 of the embodiments of the present application are described above in connection with the accompanying drawings from the perspective of functional modules. It should be understood that the functional modules can be realized in the form of hardware, or in the form of instructions of software, or in the form of a combination of hardware and software modules. Specifically, the steps of the method of determining dynamic fares for rail transit passengers crossing lines and the method embodiments of detecting in the embodiments of the present application can be completed by the integrated logic circuit of hardware in the processor and / or instructions in the form of software, and the steps of the method of determining dynamic fares for rail transit passengers crossing lines and the method of detecting disclosed in the embodiments of the present application can be directly embodied as hardware code processor execution completion, or executed by a combination of hardware and software modules in the code processor. Alternatively, the software module can be located in a mature storage medium in the art, such as random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, register, etc. The storage medium is located in the memory, and the processor reads the information in the memory, and combines the hardware to complete the steps in the method embodiments of determining dynamic fares for rail transit passengers crossing lines and the method of detecting.

[0123] Figure 6 is a schematic block diagram of a federation node layer 110 of one embodiment provided by the present application.

[0124] As shown in Figure 6 , the federation node layer 110 can include: a memory 111 and a processor 112, the memory 111 being used to store a computer program and transmit the program code to the processor 112. In other words, the processor 112 can call and run the computer program from the memory 111 to implement the method in the embodiments of the present application.

[0125] For example, the processor 112 can be used to execute the above-mentioned method embodiments according to the instructions in the computer program.

[0126] In some embodiments of the present application, the federation node layer 110 can include but is not limited to: general processor, a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field Programmable Gate Array (FPGA) or other programmable logic device, discrete gate or transistor logic, discrete hardware components, etc.

[0127] In some embodiments of the present application, the memory 111 includes, but is not limited to: volatile memory and / or non-volatile memory. The non-volatile memory can be a Read-Only Memory (ROM), a Programmable ROM (PROM), an Erasable PROM (EPROM), an Electrically EPROM (EEPROM), or a flash memory. The volatile memory can be a Random Access Memory (RAM), which is used as an external cache. By way of example, and not limitation, many forms of RAM are available, such as Static RAM (SRAM), Dynamic RAM (DRAM), Synchronous DRAM (SDRAM), Double Data Rate SDRAM (DDR SDRAM), Enhanced SDRAM (ESDRAM), Synch link DRAM (SLDRAM), and Direct Rambus RAM (DR RAM).

[0128] In some embodiments of the present application, the computer program can be divided into one or more modules, which are stored in the memory 111 and executed by the processor 112 to complete the method provided by the present application. The one or more modules can be a series of computer program instruction segments capable of completing a specific function, which are used to describe the execution process of the computer program in the controller.

[0129] As shown in Figure 6 The federal node layer 110 can also include: A transceiver 113, which can be connected to the processor 112 or the memory 111.

[0130] The processor 112 can control the transceiver 113 to communicate with other devices, specifically, can send information or data to other devices, or receive information or data sent by other devices. The transceiver 113 can include a transmitter and a receiver. The transceiver 113 can further include an antenna, and the number of antennas can be one or more.

[0131] It should be understood that the various components in the federal node layer are connected through a bus system, which includes a data bus, a power bus, a control bus, and a state signal bus in addition to the data bus.

[0132] The application also provides a central service layer, which stores a computer program, and the computer program enables the computer to execute the method of the above-mentioned method embodiment when executed by the computer. Alternatively, one embodiment of the application also provides a computer program product containing instructions, which enables the computer to execute the method of the above-mentioned method embodiment when executed by the computer.

[0133] When implemented by using software, it can be implemented in the form of a computer program product in whole or in part. The computer program product includes one or more computer instructions. When loaded and executed by a computer, the computer program instructions produce the flow or function of the embodiment of the application in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions can be stored in a computer-readable storage medium or transferred from one computer-readable storage medium to another computer-readable storage medium, for example, the computer instructions can be transferred from one website, computer, server or data center to another website, computer, server or data center through a wired (for example, coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (for example, infrared, wireless, microwave, etc.) manner. The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server, data center, etc. integrated with one or more available media sets. The available medium can be a magnetic medium (for example, a floppy disk, a hard disk, a magnetic tape), an optical medium (for example, a digital video disc (DVD)), or a semiconductor medium (for example, a solid state disk (SSD)) and the like.

[0134] Those skilled in the art can understand that the modules and algorithm steps of each example described in combination with the embodiments disclosed herein can be realized by electronic hardware or a combination of computer software and electronic hardware. Whether the functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0135] In several embodiments provided by the present application, it should be understood that the disclosed system, device and method can be implemented in other ways. For example, the device embodiments described above are merely illustrative, for example, the division of the modules is only a logical function division, and actual implementation can have another division manner, for example, multiple modules or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the coupling or direct coupling or communication connection between the displayed or discussed modules can be indirect coupling or communication connection through some interfaces, devices or modules, which can be electrical, mechanical or other forms.

[0136] The modules described as separate components can or can not be physically separated, and the components displayed as modules can or can not be physical modules, that is, they can be located in one place or distributed on multiple network units. Part or all of the modules can be selected to achieve the purpose of the embodiment according to actual needs. For example, the functional modules in each embodiment of the present application can be integrated in one processing module, or each module can be physically present separately, or two or more modules can be integrated in one module.

[0137] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto, and any skilled person in the art can easily think of changes or replacements within the technical scope disclosed in the present application, which should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A method for determining a dynamic fare for a rail transit passenger crossing a line, applied to a federal node layer, characterized in that, The method comprises: obtaining real-time operation data of a rail transit system sent by a data source layer after desensitization processing; the real-time operation data comprises service data provided by a first operator and real-time positioning data provided by a second operator; the service data comprises stored ticket record data and commuting regularity analysis data; the real-time positioning data comprises real-time GPS coordinate data of the rail transit vehicle and real-time passenger flow data of a rail transit station; analyzing the real-time operation data to obtain predicted passenger flow data of the rail transit system in a preset time period; encrypting and transmitting the predicted passenger flow data to a central service layer, so that the central service layer adjusts the ticket price of the rail transit system in the preset time period according to the predicted passenger flow data.

2. The method of claim 1, wherein, The joint plate node layer is deployed with a first edge node and a second edge node; the analyzing the real-time operation data to obtain predicted passenger flow data of the rail transit system in a preset time period comprises: predicting passenger flow data of the rail transit system in a preset time period according to the service data provided by the first operator and a first LSTM prediction model located in the first edge node to obtain first predicted passenger flow sub-data; the first LSTM prediction model is obtained by training a preset first LSTM model based on federated learning on a blockchain according to ticket price time sequence characteristics corresponding to historical operation data of the rail transit system and historical passenger flow of the rail transit system in each time period; predicting passenger flow data of the rail transit system in a preset time period according to the real-time positioning data provided by the second operator and a second LSTM prediction model located in the second edge node to obtain second predicted passenger flow sub-data; the second LSTM prediction model is obtained by training a preset second LSTM model based on federated learning on a blockchain according to passenger flow time sequence characteristics corresponding to historical operation data of the rail transit system and historical passenger flow of the rail transit system in each time period; obtaining the predicted passenger flow data according to the first predicted passenger flow sub-data and the second predicted passenger flow sub-data.

3. The method of claim 2, wherein, Before the predicting passenger flow data of the rail transit system in a preset time period according to the service data provided by the first operator and the first LSTM prediction model located in the first edge node to obtain first predicted passenger flow sub-data, the method further comprises: aligning the service data provided by the first operator in the first edge node and the real-time positioning data provided by the second operator in the second edge node according to time stamp information to obtain aligned service data and real-time positioning data; Accordingly, the predicting passenger flow data of the rail transit system in a preset time period according to the service data provided by the first operator and the first LSTM prediction model located in the first edge node to obtain first predicted passenger flow sub-data comprises: In the first edge node, the aligned service data is subjected to time sequence feature extraction, to obtain ticket price time sequence features; The ticket price time sequence features are input into the first LSTM prediction model, to obtain first predicted passenger flow sub-data of the rail transit system within a preset time period; Correspondingly, the passenger flow data of the rail transit system within a preset time period is predicted according to the real-time positioning data provided by the second operator and a second LSTM prediction model located in the second edge node, to obtain second predicted passenger flow sub-data, which comprises: In the second edge node, the aligned real-time positioning data is subjected to time sequence feature extraction, to obtain passenger flow time sequence features; The passenger flow time sequence features are input into the second LSTM prediction model, to obtain second predicted passenger flow sub-data of the rail transit system within a preset time period.

4. The method of claim 3, wherein, The real-time operation data further comprises payment flow data provided by a UnionPay payment platform and municipal external data; before the ticket price time sequence features are input into the first LSTM prediction model, to obtain first predicted passenger flow sub-data of the rail transit system within a preset time period, the method further comprises: According to timestamp information, the payment flow data and municipal external data are subjected to data alignment processing in the first edge node in time sequence, to obtain emergency event data aligned with the aligned service data; In the first edge node, the emergency event data is subjected to time sequence feature extraction, to obtain municipal external features; Correspondingly, the ticket price time sequence features are input into the first LSTM prediction model, to obtain first predicted passenger flow sub-data of the rail transit system within a preset time period, which comprises: The ticket price time sequence features and the municipal external features are input into the first LSTM prediction model, to obtain the first predicted passenger flow sub-data.

5. The method of claim 1, wherein, The predicted passenger flow data is encrypted and transmitted to a central service layer, so that the central service layer adjusts the ticket price of the rail transit system within the preset time period according to the predicted passenger flow data, which comprises: The predicted passenger flow data is subjected to gradient clipping, to obtain the predicted passenger flow data subjected to gradient processing; The predicted passenger flow data subjected to gradient processing is injected with Gaussian noise, to obtain the predicted passenger flow data subjected to enhanced differential privacy; The predicted passenger flow data subjected to enhanced differential privacy is encrypted by using homomorphic encryption gradient, to obtain the encrypted predicted passenger flow data.

6. The method of claim 1, wherein, Further comprising: After the central service layer adjusts the ticket price within the preset time period, passenger behavior feedback information sent by an application layer is received; the passenger behavior feedback information comprises card swiping interval information, path deviation information and ticket price sensitivity information; The predicted passenger flow data of the rail transit system within a preset time period is adjusted according to the passenger behavior feedback information, to update the predicted passenger flow data.

7. A method for determining a dynamic fare for a rail transit passenger crossing lines, applied to a central service layer, characterized in that, The method comprises: Acquiring predicted passenger flow data encrypted and transmitted by a federal node layer; inputting the predicted passenger flow data into a ticket price optimization model located in the central service layer to obtain an adjustment result of a ticket price of each rail transit station in the rail transit system in a preset time period; the ticket price optimization model is obtained by training local passenger flow and basic ticket price of each historical time period based on federated learning on a block chain; sending the adjustment result of the ticket price of each rail transit station in the preset time period to an application layer, so that the application layer publishes the ticket price adjusted according to the adjustment result.

8. The method of claim 7, wherein, The training method of the ticket price optimization model is as follows: obtaining model parameters of a local ticket price making model corresponding to each edge node in the federated node layer; the local ticket price making model is obtained by training local passenger flow and basic ticket price of each historical time period corresponding to the rail transit station in the edge node; generating a ticket price optimization model by aggregating the model parameters corresponding to each edge node using a federated weighted average algorithm.

9. The method of claim 7, wherein, The training method of the local ticket price making model is as follows: for each rail transit station, determining a time period type corresponding to the rail transit station in each historical time period according to local passenger flow corresponding to the rail transit station in each historical time period; the time period type includes a peak time period, a flat peak time period and an emergency mode; assigning an adjustment coefficient to a ticket price corresponding to the rail transit station in each historical time period according to the time period type corresponding to the rail transit station in each historical time period; training the adjustment coefficient and the basic ticket price corresponding to the ticket price of the rail transit station in each historical time period in each edge node to obtain the local ticket price making model.

10. A device for determining a dynamic fare for a passenger of a rail transport across lines, applied to a federal node level, characterized in that, The device comprises: a first data acquisition module configured to acquire real-time operation data of a rail transit system sent by a data source layer after desensitization processing; the real-time operation data comprises service data provided by a first operator and real-time positioning data provided by a second operator; the service data comprises stored ticketing record data and commuting regularity analysis data; the real-time positioning data comprises real-time GPS coordinate data of the rail transit vehicle and real-time passenger flow data of the rail transit station; a prediction module configured to analyze the real-time operation data to obtain predicted passenger flow data of the rail transit system in a preset time period; an encrypted transmission module configured to encrypt and transmit the predicted passenger flow data to a central service layer, so that the central service layer adjusts a ticket price of the rail transit system in the preset time period according to the predicted passenger flow data.

11. A device for determining a dynamic fare of a rail transit passenger across lines, applied to a central service layer, characterized in that, The device comprises: a second data acquisition module configured to acquire predicted passenger flow data encrypted and transmitted by a federated node layer; a ticket price adjustment module configured to input the predicted passenger flow data into a ticket price optimization model located in the central service layer to obtain an adjustment result of a ticket price of each rail transit station in the rail transit system in a preset time period; the ticket price optimization model is obtained by training local passenger flow and basic ticket price of each historical time period based on federated learning on a block chain; The adjustment result sending module is configured to send the adjustment result of the ticket price of each rail transit station in a preset time period to an application layer, so that the application layer publishes the ticket price adjusted according to the adjustment result.

12. A federal node layer characterized in that, The computer program product comprises: A processor and a memory, the memory being configured to store a computer program, and the processor being configured to invoke and run the computer program stored in the memory to execute the method of any one of claims 1-6.

13. A central service layer, characterized by A computer program product for storing a computer program, the computer program causing a computer to execute the method of any one of claims 7-9.

Images