Regional traffic flow prediction method and system

Abstract

The present invention relates to a regional traffic flow prediction method and system, comprising: obtaining historical traffic-related data for a target area; wherein the historical traffic-related data includes historical passenger flow data, weather data, and holiday data; utilizing an ensemble learning method based on the historical traffic-related data to obtain a passenger flow prediction model; and using the passenger flow prediction model to predict passenger flow data for the target area over a future period of time. This method can effectively predict and respond to passenger flow surges, avoid traffic congestion, and ensure service quality, and is of great significance to urban traffic management.

Description

Technical Field

[0001] The present invention relates to the technical field of passenger flow prediction, and in particular to a method and system for regional traffic flow prediction. Background Art

[0002] A core technical challenge facing urban traffic management is how to accurately predict and respond to exponentially growing passenger flows. Traditional linear prediction models struggle to capture sudden surges in passenger flow, leading to untimely allocation of traffic resources, congestion, and safety hazards. Exponential passenger flow growth is characterized by rapidity and intensity, triggered by complex and variable factors, including weather changes, emergencies, and holidays. This complexity makes a single prediction model unsuitable for diverse scenarios. Furthermore, the spatial distribution of passenger flow is highly dynamic, with certain areas potentially becoming hotspots within a short period of time, posing a challenge to accurate early warning. Furthermore, the real-time and adaptability of prediction models are also key issues. How can we quickly identify characteristic patterns of exponential growth within massive amounts of data and adjust prediction parameters in a timely manner to accommodate evolving passenger flow trends? Furthermore, the generation and delivery of warning information also face the challenge of timeliness. How can we issue accurate warnings at the early stages of a passenger flow surge to buy valuable time for traffic resource allocation? Addressing these issues is crucial for improving urban traffic management, ensuring public safety, and enhancing service quality.

[0003] Quality is of the essence. Summary of the Invention

[0004] The purpose of the present invention is to solve the problems existing in the above-mentioned prior art.

[0005] To achieve the above object, the present invention provides the following solutions:

[0006] A regional traffic flow prediction method, comprising:

[0007] Acquire historical traffic-related data of the target area; wherein the historical traffic-related data includes: historical passenger flow data, weather data, and holiday data;

[0008] Based on the historical traffic correlation data, a passenger flow prediction model is obtained using an ensemble learning method;

[0009] Based on the passenger flow prediction model, passenger flow data of the target area in the future period is predicted.

[0010] Optionally, obtaining historical traffic-related data of the target area includes:

[0011] Obtain historical passenger flow data, weather data, and holiday data for the target area;

[0012] Performing statistical analysis on the historical passenger flow data by region and time period to obtain the mean, variance, and peak value statistical indicators of passenger flow in each region and time period;

[0013] Based on the statistical indicators, determine the passenger flow change trend in each area and time period;

[0014] If the passenger flow in a certain area during a certain period shows an exponential growth trend, the corresponding area and period will be marked as potential peak passenger flow areas and periods;

[0015] For the potential passenger flow peak areas and time periods, time series features of passenger flow data are extracted; wherein the time series features include the variation patterns of passenger flow mean, variance and peak value at different time scales.

[0016] Optionally, obtaining the passenger flow prediction model includes:

[0017] Classify and quantify weather data according to weather conditions, and convert holiday data into binary features;

[0018] The classified and quantified weather data, the holiday data converted into binary features, and the time series features of the passenger flow data are combined as input to train the preset integrated model to obtain the passenger flow prediction model.

[0019] Optionally, training the preset ensemble model includes:

[0020] Using the merged data, the holding vector machine, long short-term memory neural network model, and convolutional neural network model are trained respectively to obtain three sub-models;

[0021] The prediction performance of each sub-model on the merged data was evaluated through the 5-fold cross-validation method, and the weight coefficient of each sub-model in the integration was determined;

[0022] Based on each sub-model after determining the weight coefficient, the passenger flow prediction model is constructed.

[0023] Optionally, based on the passenger flow prediction model, predicting passenger flow data of the target area within a period of time in the future includes:

[0024] Determining high-risk areas and high-risk periods where exponential growth in passenger flow occurs based on the output of the passenger flow prediction model;

[0025] For each high-risk area and high-risk period, set a corresponding passenger flow warning threshold;

[0026] When the predicted passenger flow in a high-risk area and high-risk period exceeds the corresponding warning threshold, a warning message will be automatically generated.

[0027] A regional traffic flow prediction system is used to implement a regional traffic flow prediction method, the system comprising: a data acquisition module, a model building module and a passenger flow prediction module;

[0028] The data acquisition module is used to obtain historical traffic-related data of the target area; wherein the historical traffic-related data includes: historical passenger flow data, weather data and holiday data;

[0029] The model building module is used to build a passenger flow prediction model based on the historical traffic related data using an integrated learning method;

[0030] The passenger flow prediction module is used to predict the passenger flow data of the target area within a period of time in the future based on the passenger flow prediction model.

[0031] Optionally, the data acquisition module includes: a data acquisition submodule and a data analysis submodule;

[0032] The data collection submodule is used to obtain historical passenger flow data, weather data and holiday data of the target area;

[0033] The data analysis submodule is used to perform statistical analysis on the historical passenger flow data according to different regions and time periods to obtain the passenger flow mean, variance and peak statistical indicators for each region and time period;

[0034] The data analysis submodule determines the passenger flow change trend of each area and time period based on the statistical indicators; if the passenger flow of a certain area and time period shows an exponential growth trend, the corresponding area and time period are marked as potential passenger flow peak areas and time periods;

[0035] The data analysis submodule extracts time series features of passenger flow data for the potential passenger flow peak areas and time periods; wherein the time series features include the variation patterns of passenger flow mean, variance and peak value at different time scales.

[0036] Optionally, the model building module includes: a data conversion submodule and a model training submodule;

[0037] The data conversion submodule is used to classify and quantify weather data according to weather conditions and convert holiday data into binary features;

[0038] The model training submodule is used to combine the classified and quantified weather data, the holiday data converted into binary features, and the time series features of the passenger flow data as input to train the preset integrated model to obtain the passenger flow prediction model.

[0039] Optionally, the model training submodule uses the merged data to train a vector machine, a long short-term memory neural network model, and a convolutional neural network model respectively to obtain three sub-models;

[0040] The prediction performance of each sub-model on the merged data was evaluated through the 5-fold cross-validation method, and the weight coefficient of each sub-model in the integration was determined;

[0041] Based on each sub-model after determining the weight coefficient, the passenger flow prediction model is constructed.

[0042] Optionally, the passenger flow prediction module determines high-risk areas and high-risk periods where exponential growth in passenger flow occurs based on the output results of the passenger flow prediction model;

[0043] For each high-risk area and high-risk period, set a corresponding passenger flow warning threshold;

[0044] When the predicted passenger flow in a high-risk area and high-risk period exceeds the corresponding warning threshold, a warning message will be automatically generated.

[0045] The beneficial effects of the present invention are:

[0046] The present invention discloses a method and system for regional traffic flow prediction. This method obtains historical passenger flow data, uses time series analysis to identify characteristic patterns of exponential passenger flow growth, and combines external factors such as weather and holidays to construct a multi-model integrated passenger flow prediction model to improve the accuracy and adaptability of the prediction. Based on the prediction results, the present invention can automatically determine high-risk areas and time periods, generate early warning information, and trigger the early scheduling of transportation resources. Through continuous monitoring and feedback mechanisms, the prediction algorithm is dynamically optimized to provide reliable decision-making support for responding to exponential passenger flow. The present invention can effectively predict and respond to passenger flow surges, avoid traffic congestion, and ensure service quality, which is of great significance to urban traffic management. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0048] Figure 1 The figure is a flow chart of a method for predicting regional traffic flow according to an embodiment of the present invention. DETAILED DESCRIPTION

[0049] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0050] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0051] like Figure 1 As shown, this embodiment proposes a regional traffic flow prediction method, which specifically includes:

[0052] Obtain traffic-related data for the target area; traffic-related data includes historical passenger flow data, weather data, and holiday data;

[0053] Based on traffic correlation data, the passenger flow prediction model is obtained using ensemble learning methods;

[0054] Based on the passenger flow prediction model, predict the passenger flow data of the target area in the future.

[0055] Specifically, in this embodiment, a regional traffic flow prediction method first obtains historical passenger flow data, weather data, and holiday data, and uses an integrated learning method to establish an exponential passenger flow prediction model; based on the output results of the exponential passenger flow prediction model, high-risk areas and high-risk time periods where passenger flow increases exponentially are determined; for each high-risk area and high-risk time period, a corresponding passenger flow warning threshold is set; when the passenger flow predicted for a certain high-risk area and high-risk time period exceeds the corresponding warning threshold, a warning message is automatically generated to trigger the early scheduling of transportation resources.

[0056] Furthermore, obtaining traffic-related data of the target area includes:

[0057] Obtain historical passenger flow data, weather data, and holiday data for the target area;

[0058] Conduct statistical analysis on historical passenger flow data by region and time period to obtain the mean, variance, and peak passenger flow statistics for each region and time period;

[0059] Determine the passenger flow trend in each area and time period based on statistical indicators;

[0060] If the passenger flow in a certain area during a certain period shows an exponential growth trend, the corresponding area and period will be marked as potential peak passenger flow areas and periods;

[0061] For potential passenger flow peak areas and time periods, the time series characteristics of passenger flow data are extracted; among them, the time series characteristics include the changing patterns of passenger flow mean, variance and peak value at different time scales.

[0062] Specifically, in this embodiment, daily passenger flow data for the past year is first obtained from the transportation operations department. Weather data for the same period, including temperature, humidity, and rainfall, is also obtained from the meteorological department. Holiday information is also extracted from the calendar. When acquiring historical and real-time passenger flow monitoring data for the target area, video surveillance systems, infrared sensors, and other equipment can be used to collect and transmit this data to a database. For example, infrared sensors can be installed at the entrance of a shopping mall. Each time a customer enters or exits, the sensors automatically record and upload the data. The system also periodically extracts historical passenger flow data from the database and aggregates and analyzes it alongside real-time data. When performing statistical analysis on this data, the system divides the data into different areas and time periods. For example, the mall can be divided into clothing, dining, and entertainment areas, and statistics are collected hourly. By calculating metrics such as the mean, variance, and peak value of passenger flow for each area and time period, the system can determine trends in passenger flow. If exponential growth in passenger flow is detected in a particular area or time period, such as a 20% hourly increase in passenger flow in the dining area during lunchtime, the system automatically marks this area and time period as a potential peak area. For these potential peaks, the system will extract the time series features of its passenger flow data, such as calculating the changes in the mean, variance and peak value of passenger flow every 15 minutes and every 30 minutes.

[0063] Furthermore, obtaining a passenger flow prediction model includes:

[0064] Classify and quantify weather data according to weather conditions, and convert holiday data into binary features;

[0065] Combine the classified and quantified weather data, the binary holiday data, and the time series features of the passenger flow data as input to train the preset ensemble model. Obtain a passenger flow prediction model.

[0066] Furthermore, training the preset integrated model includes:

[0067] Using the merged data, the holding vector machine, long short-term memory neural network model, and convolutional neural network model are trained respectively to obtain three sub-models;

[0068] The prediction performance of each sub-model on the merged data was evaluated through the 5-fold cross-validation method, and the weight coefficient of each sub-model in the integration was determined;

[0069] Based on each sub-model with determined weight coefficients, a passenger flow prediction model is constructed.

[0070] Specifically, in this embodiment, when obtaining historical data on index passenger flow, in addition to passenger flow data, it is also necessary to collect data on external factors such as weather conditions, holiday information, and records of major events. For example, the daily weather conditions of the past year can be obtained from the meteorological department and classified and quantified, such as sunny days are recorded as 1, cloudy days are recorded as 2, light rain is recorded as 3, moderate rain is recorded as 4, heavy rain is recorded as 5, etc. Holiday and major event information can be extracted from news reports and converted into binary features, that is, occurrence is recorded as 1 and non-occurrence is recorded as 0. When constructing an index passenger flow prediction model, the passenger flow, weather conditions, holidays and major event features of the past 30 days can be used as input to predict the passenger flow for the next 7 days. Through methods such as grid search, the model's hyperparameters, such as the number of hidden layers, the number of hidden units, the learning rate, etc., are optimized to improve the model's prediction performance. In the model evaluation stage, indicators such as MAE and MAPE can be used to measure the prediction error. For example, if the predicted passenger flow for the next seven days is 1000, 1200, 1500, 1800, 2000, 2200, and 2500, respectively, and the actual passenger flow is 1100, 1300, 1600, 1900, 2100, 2300, and 2600, then the MAE is 100 and the MAPE is 2%. By analyzing the impact of external factors on the forecast results, we can find that holidays and major events significantly increase passenger flow, while inclement weather can reduce passenger flow. Based on this, we can further optimize feature engineering methods, such as introducing the mean feature of the week before and after holidays, to improve the robustness of the prediction model. After multiple rounds of iterative optimization, we ultimately obtained an accurate and robust exponential passenger flow prediction model, which can provide important data support for business planning, traffic scheduling, and other applications.

[0071] More specifically, this embodiment utilizes an ensemble learning approach to construct multiple different types of passenger flow prediction sub-models based on the needs of passenger flow forecasting. Based on the characteristics of passenger flow data, algorithms such as support vector machine (SVM), long short-term memory neural network (LSTM), and convolutional neural network (CNN) were selected and trained to produce three sub-models. The prediction performance of each sub-model on the test set was evaluated using a 5-fold cross-validation method, resulting in a mean absolute percentage error (MAPE) of 8% for the SVM model, 6% for the LSTM model, and 7% for the CNN model. Based on the prediction accuracy of each sub-model, the weight coefficients for the SVM, LSTM, and CNN models in the ensemble were determined to be 3, 5, and 2, respectively. The prediction results of each sub-model were weighted and fused according to the weight coefficients to produce the final passenger flow forecast. A sliding window of 30 days was used, with training data updated weekly to retrain each sub-model. The fusion weights were then updated using the most recent week's passenger flow data to adapt to dynamic changes in passenger flow patterns. By integrating multiple sub-models of different types and comprehensively utilizing the nonlinear fitting ability of the SVM model, the advantage of the LSTM model in capturing time series dependencies, and the ability of the CNN model to extract local features, the robustness and accuracy of the overall prediction are improved. In practical applications, the passenger flow prediction error is controlled within 5%, providing a reliable basis for passenger flow scheduling optimization.

[0072] Furthermore, based on the passenger flow prediction model, the passenger flow data of the target area in the future period is predicted to include:

[0073] Based on the output of the passenger flow forecasting model, identify high-risk areas and high-risk periods where passenger flow is expected to increase exponentially;

[0074] Set corresponding passenger flow warning thresholds for each high-risk area and high-risk period;

[0075] When the predicted passenger flow in a high-risk area and high-risk period exceeds the corresponding warning threshold, a warning message will be automatically generated.

[0076] Specifically, in this embodiment, the model predicts hourly passenger flow in each area for the coming week. Using anomaly detection algorithms such as IsolationForest, it identifies high-risk areas and time periods with exponential passenger flow growth trends. For each high-risk area and time period, a passenger flow warning threshold is set based on historical data for the same period. When the predicted passenger flow exceeds the threshold, a warning message is automatically generated and sent to the transportation dispatch system. The dispatch system, combined with real-time passenger flow monitoring data, uses reinforcement learning algorithms such as DQN to dynamically adjust transportation capacity allocation, increasing capacity in high-risk areas and time periods. In key areas with high passenger flow, video analysis and infrared sensors are used to count the number of people entering and exiting the station in real time. If real-time passenger flow data indicates a surge exceeding a predetermined threshold, such as exceeding 1,000 passengers per 5 minutes, the emergency plan is immediately activated. The intelligent dispatch system operates temporary bus routes to evacuate passengers and increases security personnel to maintain order. Finally, cross-validation methods are used to evaluate the accuracy of passenger flow warnings and the effectiveness of dispatch measures. Methods such as Bayesian optimization are used to automatically adjust model hyperparameters to continuously optimize prediction performance and dispatch strategies.

[0077] Furthermore, this embodiment also continuously monitors the deviation between actual passenger flow and the predicted value, dynamically adjusts the parameters and structure of the prediction model through a feedback mechanism, continuously optimizes and improves the prediction algorithm, improves the real-time and accuracy of the prediction, and provides more reliable decision-making support for responding to exponential passenger flow.

[0078] Real-time passenger flow data is obtained and compared with the predicted value generated by the prediction model to calculate the deviation value; based on the calculated deviation value, it is determined whether the deviation exceeds the preset threshold. If it exceeds the threshold, the dynamic optimization of the prediction model is triggered; the gradient descent algorithm is used to adjust the parameters of the prediction model based on the deviation value, and through multiple rounds of iterative optimization, the deviation between the predicted value and the actual passenger flow is minimized; in the model optimization process, a time decay factor is introduced to give higher weight to recent data, thereby improving the model's sensitivity and real-time performance to passenger flow changes; the optimized prediction model regenerates the passenger flow prediction value for a period of time in the future and continuously compares it with real-time data to form a dynamic feedback loop; when the deviation between the predicted value and the actual passenger flow continues to decrease and remains within an acceptable range, it is determined that the prediction model has reached a better state and the optimization can be suspended; the optimized prediction results are output to provide accurate and reliable decision-making basis for responding to possible passenger flow peaks, and assist relevant departments in formulating emergency plans in advance.

[0079] Specifically, real-time passenger flow data is collected every five minutes and compared with the predicted values ​​generated by the prediction model, and the deviation between the two is calculated. If the deviation exceeds a preset 20% threshold, dynamic optimization of the prediction model is triggered. A gradient descent algorithm is used, with the deviation value serving as the loss function, and the loss function is minimized by adjusting the model parameters. In each iteration, the model parameters are updated at a learning rate of 0.1 to 0.5 based on the deviation value. A time decay factor is introduced, assigning a weight of 5 to data within the last hour, a weight of 2 to data between 1 and 2 hours, and a weight of 1 to data older than 2 hours. After 50 iterations of optimization, the updated prediction model regenerates passenger flow forecasts for the next two hours and continuously compares them with real-time data. When the deviation between the predicted value and actual passenger flow is reduced to within 10% for three consecutive cycles, the prediction model is considered to have reached a relatively optimal state, and optimization can be paused. The optimized prediction results are output and combined with the maximum capacity and carrying capacity of the station. If the predicted passenger flow exceeds the threshold, an early warning message will be sent to the subway dispatching center 30 minutes in advance, suggesting that the number of trains be increased in advance, and security personnel will be coordinated to be prepared. If necessary, the emergency plan will be activated to evacuate stranded passengers and ensure safe and orderly operations.

[0080] This embodiment also proposes a regional traffic flow prediction system, including: a data acquisition module, a model building module and a passenger flow prediction module;

[0081] A data collection module is used to obtain historical traffic-related data of the target area; wherein the historical traffic-related data includes: historical passenger flow data, weather data, and holiday data;

[0082] The model building module is used to build a passenger flow prediction model based on historical traffic correlation data using ensemble learning methods;

[0083] The passenger flow prediction module is used to predict the passenger flow data of the target area in the future based on the passenger flow prediction model.

[0084] Furthermore, the data acquisition module includes: a data acquisition submodule and a data analysis submodule;

[0085] The data collection submodule is used to obtain historical passenger flow data, weather data, and holiday data of the target area;

[0086] The data analysis submodule is used to perform statistical analysis on historical passenger flow data by different regions and time periods, and obtain the mean, variance and peak statistical indicators of passenger flow in each region and time period;

[0087] The data analysis submodule determines the passenger flow trend in each area and time period based on statistical indicators. If the passenger flow in a certain area and time period shows an exponential growth trend, the corresponding area and time period will be marked as a potential peak passenger flow area and time period.

[0088] The data analysis submodule extracts the time series characteristics of passenger flow data for potential peak passenger flow areas and time periods. The time series characteristics include the changing patterns of passenger flow mean, variance and peak value at different time scales.

[0089] Furthermore, the model building module includes: a data conversion submodule and a model training submodule;

[0090] The data conversion submodule is used to classify and quantify weather data according to weather conditions and convert holiday data into binary features;

[0091] The model training submodule is used to combine the classified and quantified weather data, the holiday data converted into binary features, and the time series features of the passenger flow data as input, train the preset integrated model, and obtain the passenger flow prediction model.

[0092] Furthermore, the model training submodule uses the merged data to train the vector machine, long short-term memory neural network model and convolutional neural network model respectively to obtain three sub-models;

[0093] The prediction performance of each sub-model on the merged data was evaluated through the 5-fold cross-validation method, and the weight coefficient of each sub-model in the integration was determined;

[0094] Based on each sub-model after determining the weight coefficient, a passenger flow prediction model is constructed.

[0095] Furthermore, the passenger flow prediction module determines high-risk areas and high-risk periods where exponential growth in passenger flow occurs based on the output of the passenger flow prediction model;

[0096] Set corresponding passenger flow warning thresholds for each high-risk area and high-risk period;

[0097] When the predicted passenger flow in a high-risk area and high-risk period exceeds the corresponding warning threshold, a warning message will be automatically generated.

[0098] This embodiment discloses a method and system for regional traffic flow prediction. This method obtains historical and real-time passenger flow data, uses time series analysis and spatial clustering algorithms, and identifies characteristic patterns of exponential growth in passenger flow. In combination with external factors such as weather, holidays, etc., a multi-model integrated exponential passenger flow prediction model is constructed to improve the accuracy and adaptability of the prediction. Based on the prediction results, this embodiment can automatically determine high-risk areas and time periods, generate early warning information, and trigger the early scheduling of transportation resources. Through continuous monitoring and feedback mechanisms, the prediction algorithm is dynamically optimized to provide reliable decision-making support for responding to exponential passenger flow. The present invention can effectively predict and respond to passenger flow surges, avoid traffic congestion, and ensure service quality, which is of great significance to urban traffic management.

[0099] The embodiments described above are merely descriptions of preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Without departing from the spirit of the present invention, various modifications and improvements made to the technical solutions of the present invention by persons skilled in the art should fall within the scope of protection defined by the claims of the present invention.

Claims

1. A regional traffic flow prediction method, characterized in that: include: Acquire historical traffic-related data of the target area; wherein the historical traffic-related data includes: historical passenger flow data, weather data, and holiday data; Based on the historical traffic correlation data, a passenger flow prediction model is obtained using an ensemble learning method; Based on the passenger flow prediction model, predict the passenger flow data of the target area in the future period; Obtaining historical traffic-related data for the target area includes: Obtain historical passenger flow data, weather data, and holiday data for the target area; Performing statistical analysis on the historical passenger flow data by region and time period to obtain the mean, variance, and peak value statistical indicators of passenger flow in each region and time period; Based on the statistical indicators, determine the passenger flow change trend in each area and time period; If the passenger flow in a certain area during a certain period shows an exponential growth trend, the corresponding area and period will be marked as potential peak passenger flow areas and periods; Extracting time series features of passenger flow data for the potential passenger flow peak areas and time periods; wherein the time series features include the variation patterns of passenger flow mean, variance, and peak value at different time scales; Obtaining the passenger flow prediction model includes: Classify and quantify weather data according to weather conditions, and convert holiday data into binary features; The classified and quantified weather data, the holiday data converted into binary features, and the time series features of the passenger flow data are combined as inputs to train a preset integrated model to obtain the passenger flow prediction model; The passenger flow data of the target area predicted in the future period includes: Determining high-risk areas and high-risk periods where exponential growth in passenger flow occurs based on the output of the passenger flow prediction model; For each high-risk area and high-risk period, set a corresponding passenger flow warning threshold; When the predicted passenger flow in any high-risk area and high-risk period exceeds the corresponding warning threshold, a warning message will be automatically generated.

2. The regional traffic flow prediction method according to claim 1, characterized in that: Training a preset ensemble model includes: Using the merged data, the holding vector machine, long short-term memory neural network model, and convolutional neural network model are trained respectively to obtain three sub-models; The prediction performance of each sub-model on the merged data was evaluated through the 5-fold cross-validation method, and the weight coefficient of each sub-model in the integration was determined; The passenger flow prediction model is constructed based on each sub-model after the weight coefficient is determined.

3. A regional traffic flow prediction system, characterized in that: For implementing the regional traffic flow prediction method according to any one of claims 1-2, the system comprises: a data acquisition module, a model building module and a passenger flow prediction module; The data acquisition module is used to obtain historical traffic-related data of the target area; wherein the historical traffic-related data includes: historical passenger flow data, weather data and holiday data; The model building module is used to build a passenger flow prediction model based on the historical traffic related data using an integrated learning method; The passenger flow prediction module is used to predict the passenger flow data of the target area within a period of time in the future based on the passenger flow prediction model.

4. The regional traffic flow prediction system according to claim 3, characterized in that: The data acquisition module includes: a data acquisition submodule and a data analysis submodule; The data collection submodule is used to obtain historical passenger flow data, weather data and holiday data of the target area; The data analysis submodule is used to perform statistical analysis on the historical passenger flow data according to different regions and time periods to obtain the passenger flow mean, variance and peak statistical indicators for each region and time period; The data analysis submodule determines the passenger flow change trend of each area and time period based on the statistical indicators; if the passenger flow of a certain area and time period shows an exponential growth trend, the corresponding area and time period are marked as potential passenger flow peak areas and time periods; The data analysis submodule extracts time series features of passenger flow data for the potential passenger flow peak areas and time periods; wherein the time series features include the variation patterns of passenger flow mean, variance and peak value at different time scales.

5. The regional traffic flow prediction system according to claim 4, characterized in that: The model building module includes: a data conversion submodule and a model training submodule; The data conversion submodule is used to classify and quantify weather data according to weather conditions and convert holiday data into binary features; The model training submodule is used to combine the classified and quantified weather data, the holiday data converted into binary features, and the time series features of the passenger flow data as input, train the preset integrated model, and obtain the passenger flow prediction model.

6. The regional traffic flow prediction system according to claim 5, characterized in that: The model training submodule uses the merged data to train the vector machine, the long short-term memory neural network model and the convolutional neural network model respectively to obtain three sub-models; The prediction performance of each sub-model on the merged data was evaluated through the 5-fold cross-validation method, and the weight coefficient of each sub-model in the integration was determined; The passenger flow prediction model is constructed based on each sub-model after the weight coefficient is determined.

7. The regional traffic flow prediction system according to claim 3, characterized in that: The passenger flow prediction module determines high-risk areas and high-risk periods where exponential growth in passenger flow occurs based on the output results of the passenger flow prediction model; For each high-risk area and high-risk period, set a corresponding passenger flow warning threshold; When the predicted passenger flow in a high-risk area and high-risk period exceeds the corresponding warning threshold, a warning message will be automatically generated.

Images