Road traffic flow prediction method and system based on big data

Abstract

The invention relates to the technical field of traffic prediction, and discloses a road traffic flow prediction method and system based on big data, and the system comprises a data fusion processing module, a prediction algorithm matching module and a dynamic prediction execution module. The method solves the time-space sampling frequency difference between meteorological environment data and dynamic traffic flow data, achieves the refined quantitative characterization of impact factors of sudden traffic events, enhances the prediction stability under severe meteorological conditions, fuses road network adjacency relation characteristics through a time-space grid matrix, captures the cascading congestion propagation effect caused by upstream accidents, and improves the prediction accuracy. The prediction reliability of complex road network structures such as interchange junctions is improved, the congestion false alarm rate is reduced, deviation accumulation in the continuous prediction process is continuously corrected through linkage of dynamic prediction model training and a real-time feedback optimization module, the model robustness in a special traffic scene is improved, and the stability of long-term operation of a prediction system is kept.

Description

Technical Field

[0001] The present invention relates to the technical field of traffic prediction, and in particular to a road traffic flow prediction method and system based on big data. Background Art

[0002] With the acceleration of urbanization and the popularization of transportation, it has become crucial to accurately predict and fine-tune the status of urban road traffic flow. In the past few decades, with the continuous development of information technology and data mining technology, urban road traffic flow status prediction technology based on big data has also made great progress.

[0003] The current road traffic flow prediction mainly has the following technical limitations: Data silo problem: Traditional methods rely on a single data source and fail to effectively integrate multi-dimensional data such as meteorological and event data.

[0004] Static algorithm flaws: Mainstream forecasting systems rely on fixed LSTM and SVR algorithms, which are unable to dynamically adapt to changing traffic conditions. The fluctuation patterns of morning rush hour commuting traffic and holiday tourist traffic differ significantly, making it difficult for a single algorithm to address both.

[0005] 3. Spatiotemporal scale segmentation: Existing technologies simplify spatial topology into independent road segments, ignoring the cascading effects of road networks.

[0006] 4. Lack of feedback mechanism: Many commercial systems lack real-time error correction mechanisms, and the accumulation rate of continuous forecast deviation exceeds 15% / hour.

[0007] Therefore, the present invention provides a road traffic flow prediction method and system based on big data to break through the three major technical bottlenecks of dynamic fusion of multi-source heterogeneous data, intelligent adaptation of prediction algorithms, and road network spatial dependency modeling, thereby improving the prediction reliability in special scenarios and enhancing the decision-making support capabilities of intelligent transportation systems. Summary of the Invention

[0008] (1) Technical problems solved In response to the shortcomings of the existing technology, the present invention provides a road traffic flow prediction method and system based on big data, which solves the problems raised in the above background technology.

[0009] (2) Technical solution To achieve the above objectives, the present invention provides the following technical solutions: a road traffic flow prediction method and system based on big data, the method comprising the following steps: S1. Multi-source data acquisition: synchronously acquire dynamic traffic flow parameters, static environmental parameters and traffic event data; S2, spatiotemporal frequency alignment: Detect the difference in sampling frequency between dynamic and static data, and use interpolation algorithm to adjust the static data frequency to the dynamic data frequency; S3, space-time matrix construction: Based on the road network topology, the aligned data is mapped into a space-time grid matrix, which includes traffic flow, vehicle speed, environment and event characteristics; S4, algorithm intelligent matching: input the spatiotemporal grid matrix into the algorithm matching engine, and match the optimal prediction algorithm type from the preset library based on the improved Osprey optimization algorithm; S5. Dynamic model training: Call the prediction model corresponding to the matching algorithm and integrate historical patterns with real-time data for online training; S6, multi-scale prediction: output three-level traffic forecast results: short-term 15 minutes, medium-term 1 hour, and long-term 24 hours; S7. Closed-loop optimization: Dynamically adjust model parameters and algorithm matching strategies based on the deviation between prediction results and actual results.

[0010] Preferably, the S1 includes: S11. Dynamic traffic flow parameter collection: Lane-level traffic flow, average speed, and occupancy are acquired using roadside microwave detectors, checkpoint cameras, and floating vehicle GPS equipment at a sampling frequency of 5 Hz. S12. Static environmental parameter collection: temperature, precipitation, and visibility data are collected through meteorological monitoring stations, and construction closure information is obtained through the road maintenance system; S13. Extraction of sudden traffic incidents: Crawl the accident type, location coordinates and impact range from the traffic incident platform in real time, and quantify and generate the event impact factor matrix.

[0011] Preferably, the S2 includes: S21. Spatiotemporal frequency detection: Calculate the temporal frequency of dynamic traffic flow parameters and the spatial frequency of static environmental parameters ,when When triggering frequency adjustment; S22, Bilinear interpolation adjustment: Use bilinear interpolation algorithm to adjust the spatial resolution of static environmental parameters to , the interpolation formula is:

[0012] Among them, P is the interpolation result of the target position (x, y) are the grid four corner parameter values, is the position weight coefficient, and the calculation formula is in, is the boundary coordinate in the x direction, is the boundary coordinate in the y direction.

[0013] Preferably, the S3 includes: S31. Road network topology decomposition: Decompose the target road network into an m×n grid network according to the road segment nodes, with each grid covering an area of ​​500m×500m; S32, Spatiotemporal Matrix Generation: Aggregate data for each grid by time slice to generate a three-dimensional matrix .

[0014] Preferably, the S4 includes: S41. Algorithm library construction: Preset standard feature matrices of nine prediction algorithms in The historical optimal spatiotemporal grid pattern corresponding to the hth algorithm; S42. Improved Osprey matching: Initialize the osprey population position: in is the boundary of the algorithm feature space; Exploration phase update location: is the current optimal algorithm feature; Generate new locations during development: is the number of iterations; Calculate fitness: And iteratively update, output the minimum Corresponding algorithm identifier , is the real-time spatiotemporal grid matrix, is the standard characteristic matrix of the hth algorithm, Matrix Frobenius norm.

[0015] Preferably, the S3 further includes: S33. Road network topology feature fusion: The road segment connection relationship is converted into the adjacency matrix A, and the spatial dependency feature vector is generated through the graph embedding algorithm, which is then spliced ​​into the feature dimension f of the space-time grid matrix.

[0016] Preferably, the S5 includes: S51. Dynamic training mechanism: When the target algorithm is LSTM, the loss function is: in, is the loss function value, is the actual traffic flow at time t, Forecast traffic flow at time t, is the regularization coefficient, is the LSTM model parameter matrix, is the L2 regularization term; When the target algorithm is a spatiotemporal graph convolutional network, the adjacency matrix A and the spatiotemporal grid matrix are input into the graph convolution layer together.

[0017] Preferably, the S6 includes: S61, multi-scale output structure: Short-term prediction: output 15-minute granularity flow heat map; Medium-time prediction: outputs a 1-hour granularity road section saturation matrix; Long-term prediction: Output 24-hour granularity OD matrix change trend.

[0018] Preferably, including: Data fusion processing module: Integrates roadside sensing equipment, meteorological API interfaces, and event platform data crawlers to perform time-space frequency alignment and grid matrix construction; Prediction algorithm matching module: includes preset algorithm library, Osprey optimization engine and dynamic algorithm selector; Dynamic prediction execution module: includes an online training unit, a multi-scale prediction unit and a feedback optimization unit. The feedback optimization unit adjusts hyperparameters based on the prediction error.

[0019] Preferably, the data fusion processing module is deployed in an edge computing node to process road network data within a radius of 2 kilometers in real time; The prediction algorithm matching module runs on the cloud platform and dynamically loads the algorithm library through containerization technology; The dynamic prediction execution module includes a visualization interface that outputs a heat map to a traffic guidance screen and a navigation APP.

[0020] (3) Beneficial effects Compared with the existing technology, the present invention provides a road traffic flow prediction method and system based on big data, which has the following beneficial effects: 1. By combining spatiotemporal frequency alignment with a bilinear interpolation algorithm, the spatiotemporal sampling frequency differences between meteorological data and dynamic traffic flow data are resolved, enabling refined quantitative characterization of the factors affecting sudden traffic incidents and enhancing forecast stability under adverse weather conditions.

[0021] 2. Build an intelligent algorithm matching engine based on the Osprey optimization algorithm, automatically select the optimal prediction algorithm type according to traffic status characteristics, adapt to the essential pattern differences between commuting traffic and holiday traffic, and achieve seamless switching of prediction algorithms under different traffic scenarios.

[0022] 3. By integrating road network adjacency characteristics through a spatiotemporal grid matrix, we can capture the cascading congestion propagation effect caused by upstream accidents, improve the prediction reliability of complex road network structures such as interchange hubs, and reduce the false alarm rate of congestion.

[0023] 4. By linking dynamic prediction model training with the real-time feedback optimization module, the accumulated deviations in the continuous prediction process are continuously corrected, the robustness of the model in special traffic scenarios is improved, and the long-term stability of the prediction system is maintained. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 It is a schematic diagram of the overall system architecture of the present invention.

[0025] Figure 2 Flow chart of the method of the present invention. DETAILED DESCRIPTION

[0026] 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.

[0027] See also Figure 1-2 A road traffic flow prediction method and system based on big data, the method comprises the following steps: S1. Multi-source data acquisition: synchronously acquire dynamic traffic flow parameters, static environmental parameters and traffic event data; S2, spatiotemporal frequency alignment: Detect the difference in sampling frequency between dynamic and static data, and use interpolation algorithm to adjust the static data frequency to the dynamic data frequency; S3, space-time matrix construction: Based on the road network topology, the aligned data is mapped into a space-time grid matrix, which includes traffic flow, vehicle speed, environment and event characteristics; S4, algorithm intelligent matching: input the spatiotemporal grid matrix into the algorithm matching engine, and match the optimal prediction algorithm type from the preset library based on the improved Osprey optimization algorithm; S5. Dynamic model training: Call the prediction model corresponding to the matching algorithm and integrate historical patterns with real-time data for online training; S6, multi-scale prediction: output three-level traffic forecast results: short-term 15 minutes, medium-term 1 hour, and long-term 24 hours; S7. Closed-loop optimization: Dynamically adjust model parameters and algorithm matching strategies based on the deviation between the predicted results and the actual results.

[0028] S1 includes: S11. Dynamic traffic flow parameter collection: Lane-level traffic flow, average speed, and occupancy are acquired using roadside microwave detectors, checkpoint cameras, and floating vehicle GPS equipment at a sampling frequency of 5 Hz. S12. Static environmental parameter collection: temperature, precipitation, and visibility data are collected through meteorological monitoring stations, and construction closure information is obtained through the road maintenance system; S13. Extraction of sudden traffic incidents: Crawl the accident type, location coordinates and impact range from the traffic incident platform in real time, and quantify and generate the event impact factor matrix.

[0029] S2 includes: S21. Spatiotemporal frequency detection: Calculate the temporal frequency of dynamic traffic flow parameters and the spatial frequency of static environmental parameters ,when When triggering frequency adjustment; S22, Bilinear interpolation adjustment: Use bilinear interpolation algorithm to adjust the spatial resolution of static environmental parameters to , the interpolation formula is: Among them, P is the interpolation result of the target position (x, y) are the grid four corner parameter values, is the position weight coefficient, and the calculation formula is in, is the boundary coordinate in the x direction, is the boundary coordinate in the y direction.

[0030] S3 includes: S31. Road network topology decomposition: Decompose the target road network into an m×n grid network according to the road segment nodes, with each grid covering an area of ​​500m×500m; S32, Spatiotemporal Matrix Generation: Aggregate data for each grid by time slice to generate a three-dimensional matrix .

[0031] S4 includes: S41. Algorithm library construction: Preset standard feature matrices of nine prediction algorithms in The historical optimal spatiotemporal grid pattern corresponding to the hth algorithm; S42. Improved Osprey matching: Initialize the osprey population position: in is the boundary of the algorithm feature space; Exploration phase update location: is the current optimal algorithm feature; Generate new locations during development: is the number of iterations; Calculate fitness: And iteratively update, output the minimum Corresponding algorithm identifier , is the real-time spatiotemporal grid matrix, is the standard characteristic matrix of the hth algorithm, Matrix Frobenius norm.

[0032] S3 also includes: S33. Road network topology feature fusion: The road segment connection relationship is converted into the adjacency matrix A, and the spatial dependency feature vector is generated through the graph embedding algorithm, which is then spliced ​​into the feature dimension f of the space-time grid matrix.

[0033] S5 includes: S51. Dynamic training mechanism: When the target algorithm is LSTM, the loss function is: in, is the loss function value, is the actual traffic flow at time t, Forecast traffic flow at time t, is the regularization coefficient, is the LSTM model parameter matrix, is the L2 regularization term; When the target algorithm is a spatiotemporal graph convolutional network, the adjacency matrix A and the spatiotemporal grid matrix are input into the graph convolution layer together.

[0034] S6 includes: S61, multi-scale output structure: Short-term prediction: output 15-minute granularity flow heat map; Medium-time prediction: outputs a 1-hour granularity road section saturation matrix; Long-term prediction: Output 24-hour granularity OD matrix change trend.

[0035] include: Data fusion processing module: Integrates roadside sensing equipment, meteorological API interfaces, and event platform data crawlers to perform time-space frequency alignment and grid matrix construction; Prediction algorithm matching module: includes preset algorithm library, Osprey optimization engine and dynamic algorithm selector; Dynamic prediction execution module: includes an online training unit, a multi-scale prediction unit and a feedback optimization unit. The feedback optimization unit adjusts hyperparameters based on the prediction error.

[0036] The data fusion processing module is deployed on the edge computing node to process the road network data within a radius of 2 kilometers in real time; The prediction algorithm matching module runs on the cloud platform and dynamically loads the algorithm library through containerization technology; The dynamic prediction execution module includes a visualization interface that outputs a heat map to a traffic guidance screen and a navigation APP.

[0037] Example 1: Holiday highway traffic warning scenario Innovative implementation of data fusion: Synchronously access multi-source real-time data streams through the provincial traffic big data platform, specifically including: dynamic traffic flow parameters are obtained from the ETC gantry system at a frequency of 5Hz to obtain cross-sectional traffic volume and vehicle type distribution; static environmental parameters are integrated with visibility warnings and road icing red alerts issued by the Meteorological Bureau; traffic event data is used to extract accident information from the highway traffic police command platform in real time; in the spatiotemporal frequency alignment processing, in view of the sampling frequency difference between meteorological data and traffic flow data, a three-dimensional linear interpolation algorithm is used to increase the spatial resolution of visibility data to 500 meters / 5Hz, and a spatiotemporal grid matrix containing traffic mutation coefficient, visibility attenuation factor, and accident impact radius is generated. In the road network topology modeling stage, the provincial highway network is used as the benchmark to decompose the road network into 328 topological units, each of which contains the characteristic vector of the linkage relationship between the main line, ramp and service area.

[0038] Dynamic Algorithm Matching Mechanism: When the Osprey Optimization Engine detects peak holiday travel patterns, a multi-stage algorithm matching process is triggered. During the exploration phase, random search is used to identify the basic framework of the spatiotemporal graph convolutional network. During the development phase, an attention mechanism module is added to generate the algorithm identifier G=STGCN-ATT. The model training process utilizes a bidirectional transfer learning mechanism, incorporating historical travel patterns from similar holidays over the past three years and performing online training using a dynamic learning rate adjustment strategy. The multi-scale prediction module outputs three levels of warning: a short-term prediction of a 5-kilometer queue behind the accident site in 30 minutes; a medium-term prediction of a peak road network saturation of 92% within two hours; and a long-term prediction of a 24-hour interprovincial traffic volume exceeding the design capacity by 40%.

[0039] Example 2: Tidal traffic prediction scenario for urban commuting corridors Implementation of multi-source heterogeneous data fusion: To address the tidal traffic characteristics of weekday commuting corridors, the system synchronously collects multi-dimensional data through the city's intelligent transportation cloud platform. Dynamic traffic flow parameters are obtained from underground coil detectors at a 4Hz frequency, using lane-level traffic density and average speed. Static environmental parameters are integrated with subway card swipe data and shared bicycle GPS heat maps. Traffic event data is connected to the traffic police signal control system in real time. During the spatiotemporal frequency alignment phase, a time series resampling algorithm is used to align commuting demand data to the traffic flow sampling frequency, addressing the time-frequency differences between subway card swipe data and traffic flow data. This creates a fusion feature matrix that includes tidal factors, bus arrival frequency, and subway transfer intensity. Road network topology modeling decomposes the commuting corridor into a three-level dynamic grid: core area, transition area, and radial area. Lane direction weight parameters are embedded in each grid.

[0040] The Osprey optimization algorithm dynamically adapts: When tidal traffic characteristics are detected during the morning rush hour, the algorithm matching engine initiates targeted optimization. During the exploration phase, a gradient search is used to lock in the graph neural network framework. During the development phase, a spatiotemporal attention mechanism is integrated to generate the algorithm identifier G=GNN-TSA. Dynamic model training utilizes an incremental learning mechanism, loading a library of historical tidal traffic patterns and updating model parameters in real time through adaptive batch processing. The multi-scale prediction module outputs graded warnings: Short-term predictions indicate that queues on the main access roads to the city will extend to the upstream interchange in 15 minutes; medium-term predictions indicate that the core area saturation will reach the critical value of 95% within one hour; and long-term predictions indicate that the delay index for outbound traffic during the evening rush hour will rise to 2.8.

[0041] Closed-loop optimization and traffic guidance: Based on electronic police video streams and bus arrival information, the system detects prediction deviations in real time. The optimization module implements a dual-strategy closed-loop: when the prediction saturation error exceeds 8%, the neighbor node aggregation depth of the GNN model is automatically enhanced; when a sudden weather change is detected, the algorithm strategy is immediately switched to a combined prediction; the traffic guidance system is simultaneously activated: the tidal lane signal scheme switches phase 10 minutes in advance; the bus dispatch center increases the frequency of express trains at major stations; and the on-board navigation system promotes alternative routes such as "detour parallel branches." After a week of field testing and verification, the system has increased the average travel speed of the commuter corridor during the morning rush hour by 33%, and the tidal lane utilization rate has reached 92% of the design value.

[0042] Closed-Loop Optimization and Emergency Response: Based on drone patrol video and data from mobile police terminals, the system corrects prediction errors in real time. The optimization module triggers a three-level response: upon first detecting a queue exceeding 3 kilometers, it automatically adjusts the model's convolution kernel parameters and enhances the weight of the accident point's features. When the prediction saturation exceeds the 90% threshold, the algorithm matching strategy switches to a combined prediction mode. The emergency response system simultaneously outputs control plans: the mainline information board issues a "accident ahead, speed limit 60" command; the ramp control system activates intermittent toll booth releases; and the navigation platform pushes an alternative route, "detour via the G50S Shanghai-Chongqing South Line." Field-proven, the system improved traffic efficiency at the accident point by 47% within two hours of activation and reduced the duration of network-level congestion to 65% of the original forecast.

[0043] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply the existence of any such actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article, or device comprising a series of elements includes not only those elements, but also other elements not explicitly listed, or elements inherent to such process, method, article, or device. In the absence of further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of other identical elements in the process, method, article, or device comprising the element.

[0044] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A road traffic flow prediction method based on big data, characterized by: The method comprises the following steps: S1. Multi-source data acquisition: synchronously acquire dynamic traffic flow parameters, static environmental parameters and traffic event data; S2, spatiotemporal frequency alignment: Detect the difference in sampling frequency between dynamic and static data, and use interpolation algorithm to adjust the static data frequency to the dynamic data frequency; S3, space-time matrix construction: Based on the road network topology, the aligned data is mapped into a space-time grid matrix, which includes traffic flow, vehicle speed, environment and event characteristics; S4, algorithm intelligent matching: input the spatiotemporal grid matrix into the algorithm matching engine, and match the optimal prediction algorithm type from the preset library based on the improved Osprey optimization algorithm; S5. Dynamic model training: Call the prediction model corresponding to the matching algorithm and integrate historical patterns with real-time data for online training; S6, multi-scale prediction: output three-level traffic forecast results: short-term 15 minutes, medium-term 1 hour, and long-term 24 hours; S7. Closed-loop optimization: Dynamically adjust model parameters and algorithm matching strategies based on the deviation between prediction results and actual results.

2. The method for predicting road traffic flow based on big data according to claim 1, characterized in that: Said S1 comprises: S11. Dynamic traffic flow parameter collection: Lane-level traffic flow, average speed, and occupancy are acquired using roadside microwave detectors, checkpoint cameras, and floating vehicle GPS equipment at a sampling frequency of 5 Hz. S12. Static environmental parameter collection: temperature, precipitation, and visibility data are collected through meteorological monitoring stations, and construction closure information is obtained through the road maintenance system; S13. Extraction of sudden traffic incidents: Crawl the accident type, location coordinates and impact range from the traffic incident platform in real time, and quantify and generate the event impact factor matrix.

3. The method for predicting road traffic flow based on big data according to claim 1, characterized in that: The S2 includes: S21. Spatiotemporal frequency detection: Calculate the temporal frequency of dynamic traffic flow parameters and the spatial frequency of static environmental parameters ,when When triggering frequency adjustment; S22, Bilinear interpolation adjustment: Use bilinear interpolation algorithm to adjust the spatial resolution of static environmental parameters to , the interpolation formula is: Among them, P is the interpolation result of the target position (x, y) are the grid four corner parameter values, is the position weight coefficient, and the calculation formula is in, is the boundary coordinate in the x direction, is the boundary coordinate in the y direction.

4. The method for predicting road traffic flow based on big data according to claim 1, characterized in that: The S3 includes: S31. Road network topology decomposition: Decompose the target road network into an m×n grid network according to the road segment nodes, with each grid covering an area of ​​500m×500m; S32, Spatiotemporal Matrix Generation: Aggregate data for each grid by time slice to generate a three-dimensional matrix .

5. The method for predicting road traffic flow based on big data according to claim 1, characterized in that: The S4 includes: S41. Algorithm library construction: Preset standard feature matrices of nine prediction algorithms in The historical optimal spatiotemporal grid pattern corresponding to the hth algorithm; S42. Improved Osprey matching: Initialize the osprey population position: in is the boundary of the algorithm feature space; Exploration phase update location: is the current optimal algorithm feature; Generate new locations during development: is the number of iterations; Calculate fitness: And iteratively update, output the minimum Corresponding algorithm identifier , is the real-time spatiotemporal grid matrix, is the standard characteristic matrix of the hth algorithm, Matrix Frobenius norm.

6. The method for predicting road traffic flow based on big data according to claim 1, characterized in that: Said S3 further comprises: S33. Road network topology feature fusion: The road segment connection relationship is converted into the adjacency matrix A, and the spatial dependency feature vector is generated through the graph embedding algorithm, which is then spliced ​​into the feature dimension f of the space-time grid matrix.

7. The method for predicting road traffic flow based on big data according to claim 1, characterized in that: The S5 includes: S51. Dynamic training mechanism: When the target algorithm is LSTM, the loss function is: in, is the loss function value, is the actual traffic flow at time t, Forecast traffic flow at time t, is the regularization coefficient, is the LSTM model parameter matrix, is the L2 regularization term; When the target algorithm is a spatiotemporal graph convolutional network, the adjacency matrix A and the spatiotemporal grid matrix are input into the graph convolution layer together.

8. The method for predicting road traffic flow based on big data according to claim 1, characterized in that: The S6 includes: S61, multi-scale output structure: Short-term prediction: output 15-minute granularity flow heat map; Medium-time prediction: outputs a 1-hour granularity road section saturation matrix; Long-term prediction: Output 24-hour granularity OD matrix change trend.

9. A road traffic flow prediction system based on big data, used to implement the method according to any one of claims 1 to 8, characterized in that: include: Data fusion processing module: Integrates roadside sensing equipment, meteorological API interfaces, and event platform data crawlers to perform time-space frequency alignment and grid matrix construction; Prediction algorithm matching module: includes preset algorithm library, Osprey optimization engine and dynamic algorithm selector; Dynamic prediction execution module: includes an online training unit, a multi-scale prediction unit and a feedback optimization unit. The feedback optimization unit adjusts hyperparameters based on the prediction error.

10. The road traffic flow prediction system based on big data according to claim 9, characterized in that: The data fusion processing module is deployed on the edge computing node to process the road network data within a radius of 2 kilometers in real time; The prediction algorithm matching module runs on the cloud platform and dynamically loads the algorithm library through containerization technology; The dynamic prediction execution module includes a visualization interface that outputs a heat map to a traffic guidance screen and a navigation APP.

Images