A traffic health diagnosis system to improve the accuracy of target problem identification at intersections
By combining big data and artificial intelligence technology, integrating multi-dimensional traffic data, building an XGBoost model for judging the target problem in intersections, the problem of insufficient accuracy in target problem identification in traffic health assessment in the existing technology is solved, and more accurate traffic health assessment and real-time decision support is achieved.
Patent Information
- Application Number
- CN202510027252.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-01-08
- Publication Date
- 2025-04-08
AI Technical Summary
The prior art is difficult to accurately evaluate traffic health status, especially in the identification of intersection target problems, where there is a problem of insufficient accuracy.
A traffic health diagnosis system combining big data and artificial intelligence technology is adopted. Through data collection, preprocessing, feature extraction and selection, model training and prediction evaluation, environmental factors, traffic flow, road conditions, traffic events and driving behavior data are integrated to construct an XGBoost model to determine intersection target problems.
It significantly improves the accuracy of judging target problems, provides real-time decision-making support, helps traffic management departments to timely discover and solve bottleneck problems in the traffic system, optimizes traffic resource allocation, and improves the overall operating efficiency and service level of the traffic system.
Smart Images
Figure CN119785593A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of traffic health assessment, and in particular to a traffic health diagnosis system for improving the accuracy of identifying target problems at intersections. Background Art
[0002] With the acceleration of modern urbanization, urban transportation systems are facing unprecedented challenges of complexity and diversity. The surge in traffic volume, the continuous expansion of road networks, the diversity of traffic participants' behaviors, and the growing travel demand together constitute a major challenge for urban traffic management. In this context, accurately assessing the health of traffic, that is, comprehensively and in real time understanding the operating efficiency, safety, and service quality of the traffic system, has become a focus of common concern for traffic management departments and intelligent transportation system developers.
[0003] The importance of traffic health assessment is self-evident. It is not only the basis for urban traffic planning and decision-making, but also the key to achieving the goals of easing traffic congestion, reducing traffic accidents, improving traffic efficiency, and improving the traffic environment. Through accurate traffic health assessment, traffic managers can promptly identify and solve bottleneck problems in the traffic system, optimize traffic resource allocation, and improve the overall operating efficiency and service level of the traffic system.
[0004] Therefore, with the increasing complexity of urban traffic, how to accurately evaluate the health of traffic has become a key issue that needs to be solved in current traffic management and intelligent transportation system optimization. This requires us to make full use of modern scientific and technological means, especially big data and artificial intelligence technology, combined with multi-dimensional traffic data, to conduct in-depth research and practical exploration to promote the sustainable development of urban transportation systems. Summary of the invention
[0005] In view of the shortcomings of the prior art, the present invention provides a traffic health diagnosis system that improves the accuracy of target problem identification at intersections. It uses big data and artificial intelligence technology, combined with multi-dimensional traffic data, to evaluate the health of traffic intersections and provide real-time decision support based on the evaluation results. By integrating environmental factors, traffic flow, road conditions, traffic events and driving behavior data, the accuracy of target problem identification in traffic health assessment is improved.
[0006] In order to achieve the above object, the present invention adopts the following technical solutions: A traffic health diagnosis system for improving the accuracy of identifying target problems at intersections, including the following modules: Data collection module, used to collect multi-dimensional data on traffic flow, vehicle speed, traffic signals, weather, air quality, road conditions, traffic incidents and driver behavior; Data preprocessing module, used to clean, denoise, fill missing values and process outliers on the collected data; The feature extraction and selection module extracts time features, traffic flow features, environmental factors, road conditions and driving behavior features, and uses principal component analysis to reduce and select features; The model training module uses the XGBoost model for training based on the extracted features to obtain an efficient intersection target problem discrimination model; The prediction and evaluation module predicts real-time traffic data through the trained XGBoost model and outputs the intersection health evaluation results; The result output module displays the intersection health assessment results for reference and decision-making by the traffic management department; The prediction formula of the XGBoost model is:
[0007] in, For the model The predicted value at the time of the round iteration, For the The output of a tree, is the input feature.
[0008] Preferably, the feature extraction and selection module performs feature dimension reduction by principal component analysis, and the calculation process includes the following steps: The collected raw data is standardized, and the standardized data is:
[0009] in, For the The sample Features, For the The mean of the features, For the The standard deviation of the feature; Compute the covariance matrix of the standardized data :
[0010] in, is the sample size, is the standardized data matrix, yes The transposed matrix of Perform eigenvalue decomposition on the covariance matrix to obtain the eigenvalue , ,…, and the corresponding eigenvector , ,…, ; Before selection principal components and construct a new feature matrix:
[0011] in, For the front The matrix consists of eigenvectors, is the data after dimensionality reduction.
[0012] Preferably, the training process of the XGBoost model includes the following steps: Calculate the gradient and second-order gradient of each sample point, respectively: First-order gradient:
[0013] Second order gradient:
[0014] in, is the loss function; Use the first-order and second-order gradients to calculate the leaf node output value of each tree:
[0015] in, For the The sample set of leaf nodes, is the regularization parameter; The final prediction result is obtained by weighted average:
[0016] in, is the total number of trees, For the The output of the tree.
[0017] Preferably, the loss function is:
[0018] in, For the The loss function of samples is is the regularization term, specifically:
[0019] in, is the number of leaf nodes in the tree, and is the regularization parameter, is the weight of the tree.
[0020] Preferably, the model training module further includes a hyperparameter optimization module, which uses grid search or random search to tune the learning rate, tree depth, and regularization parameters, and the optimized hyperparameters are used to improve the prediction accuracy of the model.
[0021] Preferably, the evaluation and prediction module calculates the prediction error of each sample, evaluates the model using cross-validation, and calculates indicators such as accuracy, recall, and F1-score to measure model performance.
[0022] Preferably, the calculation formula of the prediction error is:
[0023] in, is the predicted value of the model, is the actual value, and the prediction error is used for model tuning.
[0024] Preferably, the model training module is trained based on historical traffic data, and is dynamically updated according to real-time data such as real-time traffic, weather, accidents, etc., to continuously optimize the model to adapt to the ever-changing traffic environment.
[0025] Compared with the prior art, the present invention has the following beneficial effects: The present invention provides a traffic health diagnosis system for improving the accuracy of target problem identification at intersections. By integrating environmental factors, traffic flow, road conditions, traffic events and driving behavior data, the health status of traffic intersections is evaluated through multi-dimensional traffic data, and real-time decision support is provided based on the evaluation results, thereby improving the accuracy of target problem identification in traffic health evaluation. BRIEF DESCRIPTION OF THE DRAWINGS
[0026] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings required for use in the embodiments will be briefly introduced below. 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 creative work.
[0027] Figure 1 It is a schematic diagram of the overall process of the present invention. DETAILED DESCRIPTION
[0028] The present invention is further described in detail below with reference to the accompanying drawings.
[0029] The following description is used to disclose the present invention so that those skilled in the art can implement the present invention. The preferred embodiments described below are only examples, and those skilled in the art can think of other obvious variations. The basic principles of the present invention defined in the following description can be used for other embodiments, variations, improvements, equivalents, and other technical solutions that do not deviate from the spirit and scope of the present invention.
[0030] See also Figure 1 , a traffic health diagnosis system for improving the accuracy of target problem identification at intersections, including the following modules: Data collection module, used to collect multi-dimensional data on traffic flow, vehicle speed, traffic signals, weather, air quality, road conditions, traffic incidents and driver behavior; Data preprocessing module, used to clean, denoise, fill missing values and process outliers on the collected data; The feature extraction and selection module extracts time features, traffic flow features, environmental factors, road conditions and driving behavior features, and uses principal component analysis to reduce and select features; The model training module uses the XGBoost model for training based on the extracted features to obtain an efficient intersection target problem discrimination model; The prediction and evaluation module predicts real-time traffic data through the trained XGBoost model and outputs the intersection health evaluation results; The result output module displays the intersection health assessment results for reference and decision-making by the traffic management department.
[0031] In order to comprehensively evaluate traffic health, the data collection module of the present invention collects and integrates the following multi-dimensional data: Traditional traffic data: Traffic flow: traffic flow information at each road section and intersection, including vehicle flow, pedestrian and non-motor vehicle flow; Vehicle speed data: average vehicle speed and speed distribution on the road section; Traffic signal information: traffic signal timing, cycle and signal status.
[0032] Environmental factors: Weather data: temperature, humidity, precipitation, wind speed, weather type, etc.; Air quality data: including the concentration of pollutants such as PM2.5, PM10, and NO2.
[0033] Road conditions: Number of lanes: the number of lanes in the road section; Pavement quality: type of pavement (such as asphalt, concrete, etc.) and degree of damage.
[0034] Traffic incidents: Traffic accident: time, location, type of accident, etc.; Construction information: construction area, construction time and the impact of construction on traffic flow.
[0035] Driving behavior data: Sudden braking and acceleration: reflects the driver's sudden braking and acceleration behaviors; Speeding and swaying: reflects the frequency of speeding and lane departure.
[0036] The data preprocessing module of the present invention is a key step to ensure the accuracy and robustness of the model by cleaning, denoising, filling missing values and processing outliers on the collected data. The following steps are used for data preprocessing:
[0037] The data cleaning unit uses the outlier detection method Z-score to remove abnormal data. According to the characteristics and requirements of the data set, a reasonable Z-score threshold is set, and outliers that exceed the threshold are deleted.
[0038] Data denoising unit: denoises data that is erroneous or deviates from expected values, using regression methods to smooth data and reduce the interference of noise on model training.
[0039] Missing value filling: For missing data, KNN interpolation or linear interpolation is used to fill it.
[0040] Outlier detection and processing: Use the Z-score method to standardize the data and identify and remove outliers.
[0041] The feature extraction and selection module of the present invention performs the following feature extraction: Time feature extraction: Extract statistical features of different time windows (such as hours, days, and weeks) based on the time series of traffic flow, vehicle speed and other data, such as mean, standard deviation, maximum value, minimum value, etc.; Traffic flow characteristics: calculate traffic density, signal cycle and other characteristics; Environmental factor characteristics: convert weather, temperature, humidity and other information into numerical features to represent the impact of the environment on traffic health; Road condition feature: Score the road surface quality of each road section and generate a "road condition score" feature based on indicators such as the number of lanes; Driving behavior characteristics: Statistics on the frequency and duration of the driver's sudden braking, speeding, swaying and other behaviors.
[0042] Furthermore, the feature extraction and selection module uses principal component analysis to perform feature dimensionality reduction and selection on the extracted time features, traffic flow features, environmental factor features, road condition features and driving behavior features.
[0043] In order to remove redundant information and improve computational efficiency, the principal component analysis (PCA) method is used to reduce the dimensionality of high-dimensional data and screen out the most representative features. PCA can achieve feature selection through the following steps:
[0044] Principal Component Analysis (PCA) Process: Data standardization: First, the raw data is standardized to ensure that the units of different features are consistent. The standardized data is:
[0045]
[0046] in, For the The sample Features, For the The mean of the features, For the The standard deviation of a feature.
[0047] Calculate the covariance matrix: The covariance matrix of the standardized data matrix Z is:
[0048] in, is the sample size, is the standardized data matrix, yes The transposed matrix of .
[0049] Eigenvalue decomposition: Perform eigenvalue decomposition on the covariance matrix to obtain the eigenvalue , ,…, and the corresponding eigenvector , ,…, , the eigenvalue reflects the variance contribution of each principal component in the data.
[0050] Select the principal component: Select the first principal components, retaining most of the variability of the data. principal components, then the new feature matrix is:
[0051]
[0052] in, For the front The matrix consists of eigenvectors, is the data after dimensionality reduction.
[0053] Principal component score: The score of each sample on the selected principal component is:
[0054] in, For the The standardized data of samples, For the front The eigenvectors of the principal components.
[0055] PCA can not only effectively reduce dimensionality, but also eliminate the correlation between features and retain the features with the strongest explanatory power for the target variable.
[0056] The model training module of the present invention selects the XGBoost (gradient boosting tree) algorithm as the core of the traffic health assessment system, and constructs the XGBoost model for training. XGBoost can effectively handle nonlinear relationships, missing data, and multi-feature interactions by optimizing the additive model to gradually fit the residuals.
[0057] XGBoost algorithm calculation process: 1. Model construction: Target variable is the traffic health score, and the prediction formula of XGBoost is:
[0058] in, For the model The predicted value at the time of the round iteration, For the The output of a tree, is the input feature.
[0059] 2. Loss function and gradient optimization: XGBoost optimizes model parameters by minimizing the weighted loss function, which is:
[0060] is the loss function, usually squared error or logarithmic loss is used.
[0061] is a regularization term to prevent overfitting. The calculation formula is:
[0062] in, is the number of leaf nodes in the tree, and is the regularization parameter, is the weight of the tree.
[0063] 3. Calculation of gradient and second-order gradient: First-order gradient:
[0064] Second order gradient:
[0065] in, is the loss function.
[0066] 4. Calculation of leaf node output value: Use first-order and second-order gradients to calculate the leaf node output value of each tree:
[0067] in, For the leaf nodes, this episode, is the regularization parameter.
[0068] 5. Model update and prediction: After each round of iteration, the final prediction result is obtained by weighted average:
[0069] in, is the total number of trees, For the The output of the tree.
[0070] 6. Hyperparameter Optimization: Use grid search or random search to tune the learning rate, tree depth, and regularization parameters, and the optimized hyperparameters are used to improve the prediction accuracy of the model.
[0071] Evaluation indicators of the prediction and evaluation module of the present invention: Precision: An overall measure of the accuracy of a model’s predictions.
[0072] Recall rate: The proportion of samples that are correctly identified by the model among all samples that are actually positive. Recall rate is particularly important in traffic safety assessment.
[0073] F1-score: It is a harmonic mean that takes into account both precision and recall, and is particularly suitable for datasets with imbalanced categories.
[0074] The evaluation and prediction module calculates the prediction error of each sample, evaluates the model using cross-validation, and calculates indicators such as accuracy, recall, and F1-score to measure model performance.
[0075] The calculation formula of the prediction error is:
[0076] in, is the predicted value of the model, is the actual value, and the prediction error is used for model tuning.
[0077] It should be noted that the model training module of the present invention can be trained based on historical traffic data, and dynamically updated according to real-time data such as real-time traffic, weather, accidents, etc., to continuously optimize the model to adapt to the ever-changing traffic environment.
[0078] The result output module of the present invention comprises: Real-time diagnostic platform: Build a traffic health assessment system based on a cloud platform, display real-time traffic flow, intersection target problems and health assessment results through a dashboard, and support decision makers to make adjustments based on real-time data.
[0079] Early warning system: When traffic health indicators are abnormal, the system automatically issues an early warning to remind traffic management personnel to take emergency measures (such as adjusting traffic lights, diverting traffic, etc.).
[0080] Those skilled in the art should understand that the embodiments of the present invention described above and shown in the accompanying drawings are only examples and do not limit the present invention. The purpose of the present invention has been fully and effectively achieved. The functions and structural principles of the present invention have been demonstrated and explained in the embodiments, and the embodiments of the present invention may be deformed or modified in any way without departing from the principles.
Claims
1. A traffic health diagnosis system for improving the accuracy of identifying target problems at intersections, characterized by: Includes the following modules: Data collection module, used to collect multi-dimensional data on traffic flow, vehicle speed, traffic signals, weather, air quality, road conditions, traffic incidents and driver behavior; Data preprocessing module, used to clean, denoise, fill missing values and process outliers on the collected data; The feature extraction and selection module extracts time features, traffic flow features, environmental factors, road conditions and driving behavior features, and uses principal component analysis to reduce and select features; The model training module uses the XGBoost model for training based on the extracted features to obtain an efficient intersection target problem discrimination model; The prediction and evaluation module predicts real-time traffic data through the trained XGBoost model and outputs the intersection health evaluation results; The result output module displays the intersection health assessment results for reference and decision-making by the traffic management department; The prediction formula of the XGBoost model is: ； in, For the model The predicted value at the time of the round iteration, For the The output of a tree, is the input feature.
2. A traffic health diagnosis system for improving the accuracy of identifying target problems at intersections according to claim 1, characterized in that: The feature extraction and selection module performs feature dimension reduction through principal component analysis, and the calculation process includes the following steps: The collected raw data is standardized, and the standardized data is: ； in, For the The sample Features, For the The mean of the features, For the The standard deviation of the feature; Compute the covariance matrix of the standardized data : ； in, is the sample size, is the standardized data matrix, yes The transposed matrix of Perform eigenvalue decomposition on the covariance matrix to obtain the eigenvalue , ,…, and the corresponding eigenvector , ,…, ; Before selection principal components and construct a new feature matrix: ； in, For the front The matrix consists of eigenvectors, is the data after dimensionality reduction.
3. A traffic health diagnosis system for improving the accuracy of identifying target problems at intersections according to claim 2, characterized in that: The training process of the XGBoost model includes the following steps: Calculate the gradient and second-order gradient of each sample point, respectively: First-order gradient: ； Second order gradient: ； in, is the loss function; Use the first-order and second-order gradients to calculate the leaf node output value of each tree: ； in, For the The sample set of leaf nodes, is the regularization parameter; The final prediction result is obtained by weighted average: ； in, is the total number of trees, For the The output of the tree.
4. A traffic health diagnosis system for improving the accuracy of identifying target problems at intersections according to claim 3, characterized in that: The loss function is: ； in, For the The loss function of samples is is the regularization term, specifically: ； in, is the number of leaf nodes in the tree, and is the regularization parameter, is the weight of the tree.
5. A traffic health diagnosis system for improving the accuracy of identifying target problems at intersections according to claim 4, characterized in that: The model training module further includes a hyperparameter optimization module, which uses grid search or random search to tune the learning rate, tree depth, and regularization parameters. The optimized hyperparameters are used to improve the prediction accuracy of the model.
6. A traffic health diagnosis system for improving the accuracy of identifying target problems at intersections according to claim 5, characterized in that: The evaluation and prediction module calculates the prediction error of each sample, evaluates the model using cross-validation, and calculates indicators such as accuracy, recall, and F1-score to measure model performance.
7. A traffic health diagnosis system for improving the accuracy of identifying target problems at intersections according to claim 6, characterized in that: The calculation formula of the prediction error is: ； in, is the predicted value of the model, is the actual value, and the prediction error is used for model tuning.
8. A traffic health diagnosis system for improving the accuracy of identifying target problems at intersections according to claim 7, characterized in that: The model training module is trained based on historical traffic data and dynamically updated according to real-time data such as real-time traffic, weather, accidents, etc., to continuously optimize the model to adapt to the ever-changing traffic environment.