Student risk prediction method based on driving behavior modeling

Abstract

The invention relates to the technical field of machine learning, and provides a student risk prediction method based on driving behavior modeling. The method comprises the following steps: acquiring a historical driving record based on big data; extracting a first mode record in the historical driving behavior mode; reading and processing the record according to a predetermined strategy to obtain a first behavior feature vector; introducing a risk assessment mechanism to analyze the feature vector to obtain a historical risk level; constructing a risk model according to the corresponding relationship between the feature vector and the risk level; dynamically monitoring real-time behavior characteristics of students, and obtaining an initial real-time risk level through model analysis; and calibrating the initial risk level according to the multi-dimensional risk factor set to obtain a target real-time risk level, and generating a dynamic risk distribution diagram of the target trainee so as to solve the problem that the behavior risk of the driving trainee in different environments and road conditions cannot be accurately evaluated by the existing driving behavior analysis method. And the real-time performance and the accuracy of risk assessment are not enough.

Description

Technical Field

[0001] The present application relates to the field of machine learning technology, and in particular to a method for predicting student risk based on driving behavior modeling. Background Art

[0002] With the development of intelligent transportation and autonomous driving technology, driving safety has gradually become an important topic in the field of traffic management and driving training. Traditional driving behavior assessment mainly relies on manual observation and a single assessment standard, which is difficult to fully reflect the actual behavior and potential risks of drivers. In addition, the driver's behavior pattern is affected by many factors, such as road conditions, traffic environment, and driver's personal habits. Traditional methods are difficult to dynamically adapt to complex and changing driving scenarios. At present, many student driving risk assessment systems still lack accurate risk and personalized assessment methods, which makes it impossible to accurately identify the risk level of students in different driving environments, making it impossible for risk prediction and management in the student training process to achieve ideal accuracy and real-time performance. Summary of the invention

[0003] This application provides a student risk prediction method based on driving behavior modeling, aiming to solve the technical problem that the existing driving behavior analysis methods cannot accurately assess the behavioral risks of driving students in different environments and road conditions, resulting in insufficient real-time and personalization of risk assessment.

[0004] In view of the above problems, the present application provides a student risk prediction method based on driving behavior modeling.

[0005] The present application provides a student risk prediction method based on driving behavior modeling, the method comprising: obtaining historical driving records based on big data, the historical driving records comprising multiple historical driving behavior pattern records; extracting a first historical driving behavior pattern record from the multiple historical driving behavior pattern records; reading a predetermined driving behavior feature vector quantization strategy, and preprocessing the first historical driving behavior pattern record according to the predetermined driving behavior feature vector quantization strategy to obtain a first historical behavior feature vector; introducing a behavior risk assessment mechanism to perform driving risk identification analysis on the first historical behavior feature vector to obtain a first historical behavior risk level; constructing a driving behavior risk model according to a first corresponding relationship between the first historical behavior feature vector and the first historical behavior risk level; dynamically monitoring to obtain target real-time behavior feature parameters of a target student, and analyzing and identifying the target real-time behavior feature parameters through the driving behavior risk model to obtain an initial real-time behavior risk level; calibrating the initial real-time behavior risk level according to a predetermined multidimensional risk factor set to obtain a target real-time behavior risk level, and generating a target dynamic risk distribution map of the target student.

[0006] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0007] The above-mentioned student risk prediction method based on driving behavior modeling obtains the first historical driving behavior pattern record by extracting representative samples from multiple historical driving behavior pattern records, and then pre-processes these behavior patterns according to a predetermined feature vector quantization strategy to convert the original driving behavior into a standardized behavior feature vector. This process can quantify complex driving behavior information for subsequent analysis. On this basis, a behavior risk assessment mechanism is introduced to conduct an in-depth analysis of the obtained historical behavior feature vectors to identify the potential risk level of the behavior. This process can not only accurately assess the risk points in the historical driving data, but also provide a reliable basis for risk warning. Subsequently, by dynamically monitoring the student's actual driving behavior, the risk level of the student's driving behavior is accurately predicted. The real-time driving data of the target students is used to obtain the current behavioral characteristics of the target students, and these characteristics are input into the constructed driving behavior risk model for analysis, so as to realize real-time tracking and analysis of the students' driving behavior, so as to timely identify potential driving risks; when the real-time risk level deviates, it will be calibrated through a multi-dimensional risk factor set, and the assessment result will be adjusted to better fit the actual driving conditions; finally, not only the real-time behavioral risk level of the target students can be generated, but also a dynamic risk distribution map can be generated to help students understand their risk distribution in different driving scenarios and provide personalized improvement suggestions; through this dynamic adjustment and real-time feedback, students can better understand and optimize their own driving behavior, improve driving safety and training effects.

[0008] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of 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 creative work.

[0010] Figure 1 The figure is a flow chart of a method for predicting student risk based on driving behavior modeling in one embodiment.

[0011] Figure 2 The figure is a flow chart of obtaining a first historical behavior feature vector in a student risk prediction method based on driving behavior modeling in one embodiment. DETAILED DESCRIPTION

[0012] The embodiments of the present application provide a student risk prediction method based on driving behavior modeling to solve the technical problem that the existing driving behavior analysis method cannot accurately assess the behavioral risks of driving students under different environments and road conditions, resulting in insufficient real-time and personalization of risk assessment.

[0013] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application.

[0014] It should be noted that the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or server that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or modules that are not explicitly listed or inherent to these processes, methods, products or devices.

[0015] Examples, such as Figure 1 As shown, the present application provides a student risk prediction method based on driving behavior modeling, the method comprising:

[0016] Step S100: Acquire historical driving records based on big data, wherein the historical driving records include a plurality of historical driving behavior pattern records.

[0017] In an embodiment of the present application, the student's historical driving records are collected from data sources such as the student's vehicle's driving recorder and the vehicle's OBD system through big data technology. These records include the student's driving behavior data at different times and in different driving environments. These data cover multiple driving modes, such as acceleration, braking, steering, speed fluctuations, etc., reflecting the student's performance in different scenarios during driving. Each historical driving record contains the student's behavioral characteristics, such as operating methods, driving habits, and responses to emergencies. These historical driving records provide rich basic information for subsequent analysis, help understand the student's driving style and potential safety hazards, and provide solid data support for establishing an accurate driving behavior risk model.

[0018] Step S200: extracting a first historical driving behavior pattern record from the plurality of historical driving behavior pattern records.

[0019] In one embodiment, a driving behavior pattern record is randomly extracted from multiple historical driving behavior data obtained as a first historical driving behavior pattern record. This first historical driving behavior pattern record may be a record of the trainee's sudden acceleration, sudden braking or unstable turning when driving at high speed. These behaviors can better reflect the trainee's performance in a specific driving situation, and provide specific analysis data for subsequent risk assessment and prediction.

[0020] Step S300: reading a predetermined driving behavior feature vector quantization strategy, and preprocessing the first historical driving behavior pattern record according to the predetermined driving behavior feature vector quantization strategy to obtain a first historical behavior feature vector.

[0021] In one embodiment, when preprocessing the first historical driving behavior pattern record, a predetermined feature vector quantization strategy is first read. This predetermined driving behavior feature vector quantization strategy is a standardized processing method developed for driving behavior data, which is used to convert complex driving behaviors into structured numerical features. The strategy defines which driving features need to be extracted (such as acceleration amplitude, braking intensity, steering angle, etc.), and how to convert these features into numerical forms that are convenient for machine learning analysis (for example, through normalization, standardization or vector encoding); then, the predetermined driving behavior feature vector quantization strategy is applied to the extracted first historical driving behavior pattern record for preprocessing. This process extracts and converts the raw data in the first historical driving behavior pattern record into a set of feature vectors with corresponding numerical values, thereby obtaining a first historical behavior feature vector, which represents the specific behavioral characteristics of the trainee in a specific driving situation, and provides a basis for subsequent risk assessment and modeling.

[0022] Further, if Figure 2 As shown, step S300 includes:

[0023] Step S310: extracting a predetermined behavior feature indicator set in the predetermined driving behavior feature quantization strategy; step S320: traversing the predetermined behavior feature indicator set in the first historical driving behavior pattern record to obtain a first historical behavior feature parameter set; step S330: extracting a first historical parameter corresponding to a first predetermined indicator in the first historical behavior feature parameter set; step S340: performing vector analysis on the first historical parameter according to the first predetermined vector strategy for the first predetermined indicator in the predetermined driving behavior feature quantization strategy to obtain a first historical label; step S350: constructing the first historical behavior feature vector based on the first historical label.

[0024] Preferably, a set of predetermined behavior feature indicator sets are extracted from the preset driving behavior feature quantization strategy, and these indicator sets include key parameters that need to be paid attention to in driving behavior analysis, such as acceleration rate, braking intensity, steering angle, vehicle speed change, etc., and define which specific driving behavior features need to be quantified and analyzed; then, according to the extracted indicator set, the first historical driving behavior pattern record is traversed, and the data related to each indicator is extracted one by one. Through this traversal, the information in the first historical driving behavior pattern record is converted into multiple feature parameters, thereby obtaining the first historical behavior feature parameter set. For example, in a certain driving process, the acceleration rate of the trainee is 0.5m / s 2 , Braking strength is 3.2m / s 2 , the steering angle is 15°, and the vehicle speed changes from 60km / h to 40km / h; then, from the extracted first historical behavior feature parameter set, the specific data corresponding to the first predetermined indicator in the predetermined behavior feature indicator set is extracted to obtain the first historical parameter, wherein the first predetermined indicator is the first feature indicator set in the predetermined behavior feature indicator set, and can be any one of the acceleration amplitude, braking intensity, steering angle, etc., depending on the order of the feature indicators in the predetermined behavior feature indicator set; after obtaining the first historical parameter, according to the first predetermined vector strategy corresponding to the first predetermined indicator recorded in the predetermined driving behavior feature quantization strategy, the first historical parameter is vectorized and analyzed to convert it into a numerical form Formula, generate a first historical label. For example, if the first historical parameter is the braking intensity, the first predetermined vector strategy can be normalization processing, that is, converting the first historical parameter to the standard range of 0 to 1 through the maximum and minimum method; finally, adding the obtained first historical parameter to the first historical behavior feature vector as a part of the first historical behavior feature vector, repeating the above process, and adding the historical labels corresponding to the remaining historical parameters to the first historical behavior feature vector in turn, thereby constructing a complete first historical behavior feature vector. This feature vector contains quantitative information of all relevant behavior patterns of the trainee during the driving process, and provides structured basic data for subsequent risk assessment and analysis.

[0025] Step S400: introducing a behavior risk assessment mechanism to perform driving risk identification analysis on the first historical behavior feature vector to obtain a first historical behavior risk level.

[0026] In one embodiment, in order to evaluate the risk level of the first historical behavior feature vector, a behavior risk assessment mechanism is introduced. The main function of this mechanism is to analyze and evaluate the first historical behavior feature vector that has been generated, identify the corresponding risk level by combining the feature vector threshold, and output the identified risk level as the first historical behavior risk level. This risk level reflects the safety risk of the student under this specific historical driving behavior, thereby providing a basis for subsequent risk prediction and student improvement.

[0027] Further, step S400 includes:

[0028] Step S410: Read the first predetermined risk behavior characteristic vector threshold; Step S420: Determine whether the first historical behavior characteristic vector is within the first predetermined risk behavior characteristic vector threshold according to the behavior risk assessment mechanism; Step S430: If so, mark the first historical behavior characteristic vector as high risk; if not, read the second predetermined risk behavior characteristic vector threshold; Step S440: Determine whether the first historical behavior characteristic vector is within the second predetermined risk behavior characteristic vector threshold according to the behavior risk assessment mechanism; Step S450: If so, mark the first historical behavior characteristic vector as medium risk; if not, mark the first historical behavior characteristic vector as low risk; Step S460: Obtain the first historical behavior risk level based on the high risk, the medium risk and the low risk.

[0029] Preferably, a first predetermined risk behavior feature vector threshold is read from the behavior risk assessment mechanism. This threshold value represents the boundary of high-risk behavior, that is, driving behavior above this threshold value is considered to be high-risk behavior, which is usually determined based on historical driving data, accident statistics and safety standards, with the purpose of identifying those potentially dangerous driving behaviors; then, the first historical behavior feature vector is compared with the preset first risk threshold, and through the comparison method in this risk assessment mechanism, it is determined whether the first historical behavior feature vector is within the requirement of the first predetermined risk behavior feature vector threshold. If the value of the feature vector exceeds the threshold, the behavior is considered to have a higher risk, and in this case, it is marked as high risk; if the first risk threshold is not met, the first historical behavior feature vector is compared with the preset first risk threshold, and the first historical behavior feature vector is judged to have a higher risk. If the predetermined risk behavior characteristic vector threshold is reached, the second predetermined risk behavior characteristic vector threshold will continue to be read for the next step of judgment; in this step, the first historical behavior characteristic vector will be compared with the second predetermined risk behavior characteristic vector threshold in the same way to determine whether the behavior is a medium-risk behavior; if the first historical behavior characteristic vector is within the threshold range, it will be identified as medium risk; if it still does not reach the second predetermined risk behavior characteristic vector threshold, it will be identified as low risk, indicating that the risk of the driving behavior is low; finally, based on the above judgment results, the determined risk level, that is, one of high risk, medium risk and low risk, will be used as the first historical behavior risk level to reflect the risk level of the historical behavior.

[0030] Step S500: construct a driving behavior risk model based on the first corresponding relationship between the first historical behavior feature vector and the first historical behavior risk level; Step S600: dynamically monitor and obtain the target real-time behavior feature parameters of the target student, and analyze and identify the target real-time behavior feature parameters through the driving behavior risk model to obtain an initial real-time behavior risk level.

[0031] In one embodiment, after obtaining the first historical behavior feature vector and the first historical behavior risk level, the first historical behavior feature vector and the first historical behavior risk level are stored in a driving behavior risk database, and all data in the driving behavior risk database will be used as training data for constructing a driving behavior risk model; subsequently, multiple historical behavior feature vectors of each historical behavior risk level are extracted from the driving behavior risk database to construct a training data set, and then a verification data set and a test data set are constructed in the same manner; thereafter, the driving behavior risk model is supervised and learned using the training data set, and the training of the driving behavior risk model is stopped when the maximum number of iterations is reached or convergence is reached on the verification data set, and the driving behavior risk model is generalized and tested using the test data set, and compared with the set expected accuracy, if the test accuracy is greater than or equal to the expected accuracy, the current driving behavior risk model is output, otherwise, the driving behavior risk model is optimized by adjusting the model structure, hyperparameters, etc. until the driving behavior risk model meets the expected accuracy requirements, thereby providing accurate risk assessment predictions for real-time driving behavior.

[0032] In one embodiment, after constructing a driving behavior risk model, the target student's driving behavior is monitored in real time, and real-time feature parameters related to the driving behavior are collected to generate target real-time behavior feature parameters, such as the current acceleration amplitude, braking intensity, steering angle, etc. These real-time data will be continuously updated to reflect the student's current driving status; then, these target real-time behavior feature parameters are input into the driving behavior risk model for analysis. The driving behavior risk model analyzes these real-time feature data and combines previously learned historical data and risk assessment criteria to determine which risk level (e.g., high risk, medium risk, or low risk) the current driving behavior belongs to; based on these analyses, an initial real-time behavior risk level will be calculated, that is, an instant assessment result of the student's current driving behavior. This initial real-time behavior risk level reflects the student's safety in the current driving situation, and serves as the basis for subsequent real-time risk monitoring and adjustment.

[0033] Further, step S600 includes:

[0034] Step S610: construct a driving behavior risk database; Step S620: extract a first data group from the driving behavior risk database, the first data group includes a first behavior feature vector with a high risk mark; Step S630: extract a second data group from the driving behavior risk database, the second data group includes a second behavior feature vector with a medium risk mark; Step S640: extract a third data group from the driving behavior risk database, the third data group includes a third behavior feature vector with a low risk mark; Step S650: form a training data group based on the first behavior feature vector with a high risk mark, the second behavior feature vector with a medium risk mark and the third behavior feature vector with a low risk mark; Step S660: perform supervised learning on the training data group based on the support vector machine principle and verify to obtain the driving behavior risk model.

[0035] Optionally, after obtaining the first historical behavior characteristic vector and the first historical behavior risk level, the same operation is performed on multiple historical driving behavior pattern records to obtain multiple historical behavior characteristic vectors and multiple historical behavior risk levels, and a driving behavior risk database is constructed through these historical behavior characteristic vectors and historical behavior risk levels; then, a first data group is extracted from the driving behavior risk database, which includes all first behavior characteristic vectors identified as high risk (i.e., the stored historical behavior characteristic vectors), and then a second data group is extracted, which includes all second behavior characteristic vectors identified as medium risk (i.e., the stored historical behavior characteristic vectors), and a third data group is extracted, which includes all third behavior characteristic vectors identified as low risk (i.e., the stored historical behavior characteristic vectors); then, the extracted first behavior characteristic vector with high risk identification, the second behavior characteristic vector with medium risk identification, and the third behavior characteristic vector with low risk identification are compared according to a preset ratio. The three lines are divided into feature vectors to form a training data group, a validation data group and a test data group, wherein the preset ratio is usually 6:2:2; after determining the training data group, the validation data group and the test data group, an appropriate kernel function is selected for SVM model training. Commonly used kernel functions include linear kernel, radial basis kernel (RBF) kernel or polynomial kernel. For most practical problems, the RBF kernel is often selected because it can handle nonlinear classification problems; after selecting the kernel function, the hyperparameters of the SVM are determined, such as the C parameter and the Gamma parameter. The C parameter controls the degree of penalty for misclassification, while the Gamma parameter affects the complexity of the decision boundary. The optimal hyperparameters can be selected by cross-validation; in the training stage, the feature vectors in the training data group and the corresponding risk labels (high risk, medium risk, low risk) are used to train the SVM model. The goal of the SVM is to find an optimal hyperplane that maximizes the separation of samples of different categories (high risk, medium risk, low risk).By maximizing the interval between categories, SVM can find classification boundaries with good generalization ability. During the training process, SVM will also automatically identify support vectors, that is, the data points closest to the decision boundary. These support vectors are crucial to determining the final classification boundary. After the training is completed, the model is preliminarily evaluated using the validation data set. The performance of the model on the validation data set is evaluated by calculating indicators such as accuracy, precision, recall, and F1 value. The validation data set is used to adjust hyperparameters to improve the performance of the model. If the performance of the validation data set is not ideal, hyperparameters such as C value and Gamma value can be adjusted to retrain the model until the optimal model parameters are obtained. When the ideal model performance is obtained on the validation set, the final model is evaluated using the test data set. The test data set is a sample that the model has never seen, so it is used to evaluate the generalization ability of the model, that is, the performance of the model in actual applications. The evaluation indicators obtained on the test data set are used to determine whether the model can accurately predict the risk of the student's driving behavior. Finally, the trained SVM model is saved as a driving behavior risk model and used in actual driving behavior monitoring to achieve real-time prediction and risk assessment.

[0036] Further, step S600 includes:

[0037] Step S670: obtaining a target real-time behavior feature vector of the target real-time behavior feature parameters according to the predetermined driving behavior feature vector quantization strategy; Step S680: inputting the target real-time behavior feature vector into the driving behavior risk model to obtain the initial real-time behavior risk level.

[0038] Preferably, after obtaining the target real-time behavior characteristic parameters, the target real-time behavior characteristic parameters are preprocessed using a predetermined driving behavior characteristic vector quantization strategy in the same manner as described above to obtain a target real-time behavior characteristic vector of the target real-time behavior characteristic parameters; subsequently, the processed target real-time behavior characteristic vector is input into a driving behavior risk model for analysis. The model calculates the initial real-time behavior risk level of the student's current driving behavior based on the standards learned from historical data. This initial risk level reflects the student's risk level in the current driving situation, helping to promptly detect potential safety hazards or bad driving behaviors.

[0039] Further, step S680 also includes:

[0040] Step S681: Obtain the first cosine similarity between the target real-time behavior feature vector and the first historical behavior feature vector; Step S682: When the first cosine similarity reaches a predetermined similarity threshold, use the first historical behavior risk level corresponding to the first historical behavior feature vector as the second initial real-time behavior risk level of the target student; Step S683: Verify the initial real-time behavior risk level with the second initial real-time behavior risk level.

[0041] Optionally, after obtaining the initial real-time behavior risk level, in order to ensure the accuracy of the initial real-time behavior risk level, the target real-time behavior feature vector is compared with the first historical behavior feature vector, and the first cosine similarity between them is calculated. Cosine similarity is a measure of the similarity between two vectors, and the value range is from -1 to 1, where 1 indicates complete similarity, 0 indicates complete dissimilarity, and -1 indicates complete opposite. By calculating the similarity between the target student's current driving behavior and historical behavior, it can be understood whether the student's current behavior is similar to certain high-risk behaviors in the past. If the calculated first cosine similarity reaches a predetermined similarity threshold, it means that the target student's current driving behavior is similar to the historical behavior. The behavior patterns are similar. At this time, the first historical behavior risk level corresponding to the first historical behavior feature vector is used as the second initial real-time behavior risk level of the target student. After obtaining the second initial real-time behavior risk level, the second initial real-time behavior risk level is compared with the initial real-time behavior risk level calculated by the driving behavior risk model. The purpose of this step is to verify the accuracy of the preliminary assessment by comparing the results of the two. If the second initial real-time behavior risk level is consistent with the initial risk level, it means that the assessment result of the model is more accurate. If there is a difference, the risk assessment can be further analyzed and adjusted by re-data collection to ensure that the risk identification of the target student's driving behavior is more accurate.

[0042] Step S700: Calibrate the initial real-time behavior risk level according to a predetermined multi-dimensional risk factor set to obtain a target real-time behavior risk level, and generate a target dynamic risk distribution map of the target student.

[0043] In one embodiment, the initial real-time behavior risk level is further calibrated according to a predetermined multidimensional risk factor set, wherein the risk factors include road conditions (such as road flatness and slope), weather conditions (such as wind speed, temperature, humidity), etc., which may have additional effects on driving behavior and thus affect driving risk assessment. Subsequently, the target real-time risk value is obtained by multiplying the value of each risk factor by the corresponding weight (determined according to historical data) and accumulating the values, and then adding the accumulated result to the initial real-time risk level. After that, the target real-time risk value is compared with a preset discrete risk threshold, and the target real-time risk value is mapped to a target real-time behavior risk level. The preset discrete risk threshold is similar to the aforementioned predetermined risk behavior feature vector threshold, and the number is three. Then, based on the target area grid set of the target driving area, risk adjustment is performed for each area grid in combination with the predetermined multidimensional risk factor set, and the adjustment result is rendered into the grid. This process is repeated until all grids are rendered. At this time, a target dynamic risk distribution map of the target student is generated. This chart presents the real-time risk changes of the student during driving, and helps analyze in which situations the student has a higher driving risk.

[0044] Further, step S700 includes:

[0045] Step S710: The predetermined multidimensional risk factor set includes a road risk factor set and an environmental risk factor set, and the road risk factor set includes road flatness, road tortuosity and road slope, and the environmental risk factor set includes wind force, wind speed, temperature and humidity.

[0046] Preferably, the predetermined multidimensional risk factor set refers to multiple influencing factors used to evaluate driving risks, which are divided into two categories: a road risk factor set and an environmental risk factor set; wherein, the road risk factor set includes road flatness, road tortuosity and road slope; the environmental risk factor set includes wind force, wind speed, temperature and humidity; these factors will work together during real-time driving, and by evaluating these multidimensional environmental and road conditions, the risks faced by the driver can be more accurately assessed, thereby calibrating the risk level of the driving behavior.

[0047] Further, step S700 includes:

[0048] Step S720: Acquire the target driving area of ​​the target student; Step S730: Divide the target driving area to obtain a target area grid set; Step S740: Collect risks for the first grid in the target area grid set according to the predetermined multidimensional risk factor set, and obtain a first risk index by variation weighting; Step S750: Use the first risk index as a weight coefficient to adjust the target real-time behavior risk level to obtain a first predicted behavior risk level of the target student in the first grid; Step S760: Render the first predicted behavior risk level to the first grid, and generate the target dynamic risk distribution map.

[0049] Preferably, the target driving area of ​​the target student is determined by GPS positioning or other geographic information systems (GIS), and the target driving area refers to the specific geographical location or area where the student is driving, which may be a closed training ground or a specific road section, a city range, such as a practice field, a simulated road or a low-speed loop, etc.; then, the target driving area is divided into a number of small grid areas to form a target area grid set, and these grids represent different small areas in the target driving area. The driving conditions in each grid can be evaluated separately, which is convenient for a detailed analysis of the risks in the area. The way of dividing the grid can be determined according to the size of the area, the type of road or the traffic conditions; then, the risks in the first grid are collected and evaluated according to the multidimensional risk factor set, and a risk index is calculated for each grid by analyzing the specific factors in the grid (such as the road conditions and weather conditions in the area). During the calculation process, different risk factor values ​​within the grid are weighted to generate a comprehensive first risk index, which reflects the risk level within the grid. A higher value indicates a greater risk, while a lower value indicates greater safety. Then, the first risk index is used as a weight coefficient to adjust the target real-time behavioral risk level of the target student, that is, the target real-time behavioral risk level is multiplied by the first risk index to obtain the first grid risk threshold, and then the first grid risk threshold is mapped by the preset grid risk threshold to determine the first predicted behavioral risk level of the target student in the first grid. Finally, the adjusted first predicted behavioral risk level is rendered onto the first grid to form a visual dynamic risk distribution map, which shows the real-time risk situation of the target student in different grid areas. In the map, the color or mark of each grid can represent different risk levels (such as red for high-risk areas, yellow for medium-risk areas, and green for low-risk areas), which helps to intuitively display the risk situation currently faced by the student. This chart can provide feedback to the driver and can also help the instructor identify high-risk areas and intervene.

[0050] In summary, the embodiments of the present application have at least the following technical effects:

[0051] The embodiment of the present application collects the historical driving behavior patterns of students through big data and extracts features to construct a driving behavior risk model. This model is obtained by supervised learning of training data using a support vector machine (SVM); by analyzing the similarity between the real-time behavior characteristics of the target students and historical data, the initial risk level is calibrated to generate the risk assessment results of the students; combined with multidimensional risk factors such as road conditions and environmental factors, the initial risk level is weighted and adjusted, and finally a dynamic risk distribution map of the target students is obtained, providing real-time risk prediction and management. These technical effects jointly solve the technical problems of insufficient real-time and personalization of risk assessment due to the inability of existing driving behavior analysis methods to accurately assess the behavioral risks of driving students under different environments and road conditions, and realize dynamic risk assessment and optimization based on historical and real-time data, automatically identify the driving behavior risk level of students, and improve the accuracy and safety of risk assessment.

[0052] It should be noted that the above-mentioned sequence of the embodiments of the present application is only for description and does not represent the advantages and disadvantages of the embodiments. And the above-mentioned specific embodiments of this specification are described. The processes depicted in the accompanying drawings do not necessarily require the specific order and continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0053] The above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

[0054] This specification and the drawings are merely exemplary illustrations of the present application and are deemed to cover any and all modifications, variations, combinations or equivalents within the scope of the present application. Obviously, those skilled in the art may make various modifications and variations to the present application without departing from the scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the present application and its equivalents, the present application intends to include these modifications and variations.

Claims

1. A student risk prediction method based on driving behavior modeling, characterized in that: include: Acquire historical driving records based on big data, wherein the historical driving records include multiple historical driving behavior pattern records; extracting a first historical driving behavior pattern record from among the plurality of historical driving behavior pattern records; Reading a predetermined driving behavior feature quantization strategy, and preprocessing the first historical driving behavior pattern record according to the predetermined driving behavior feature quantization strategy to obtain a first historical behavior feature vector; Introducing a behavior risk assessment mechanism to perform driving risk identification analysis on the first historical behavior feature vector to obtain a first historical behavior risk level; constructing a driving behavior risk model according to a first corresponding relationship between the first historical behavior feature vector and the first historical behavior risk level; Dynamically monitoring and obtaining target real-time behavior characteristic parameters of the target trainee, and analyzing and identifying the target real-time behavior characteristic parameters through the driving behavior risk model to obtain an initial real-time behavior risk level; The initial real-time behavior risk level is calibrated according to a predetermined multi-dimensional risk factor set to obtain a target real-time behavior risk level, and a target dynamic risk distribution map of the target student is generated.

2. The student risk prediction method based on driving behavior modeling according to claim 1 is characterized in that: Reading a predetermined driving behavior feature vectorization strategy, and preprocessing the first historical driving behavior pattern record according to the predetermined driving behavior feature vectorization strategy to obtain a first historical behavior feature vector, including: Extracting the predetermined driving behavior feature into a predetermined behavior feature index set in a quantization strategy; Traversing the predetermined behavior characteristic index set in the first historical driving behavior pattern record to obtain a first historical behavior characteristic parameter set; Extracting a first historical parameter corresponding to a first predetermined indicator from the first historical behavior characteristic parameter set; Performing vector analysis on the first historical parameter according to a first predetermined vector strategy for the first predetermined indicator in the predetermined driving behavior feature quantization strategy to obtain a first historical label; The first historical behavior feature vector is constructed based on the first historical tag.

3. The student risk prediction method based on driving behavior modeling according to claim 1 is characterized in that: A behavior risk assessment mechanism is introduced to perform driving risk identification analysis on the first historical behavior feature vector to obtain a first historical behavior risk level, including: Reading a first predetermined risk behavior feature vector threshold; Determining whether the first historical behavior feature vector is within the first predetermined risk behavior feature vector threshold according to the behavior risk assessment mechanism; If so, the first historical behavior feature vector is marked as high risk; if not, a second predetermined risk behavior feature vector threshold is read; Determining whether the first historical behavior feature vector is within the second predetermined risk behavior feature vector threshold according to the behavior risk assessment mechanism; If yes, then the first historical behavior feature vector is marked as medium risk; if no, then the first historical behavior feature vector is marked as low risk; The first historical behavior risk level is obtained based on the high risk, the medium risk, and the low risk.

4. The student risk prediction method based on driving behavior modeling according to claim 3 is characterized in that: Dynamically monitor and obtain target real-time behavior characteristic parameters of the target student, and analyze and identify the target real-time behavior characteristic parameters through the driving behavior risk model to obtain an initial real-time behavior risk level, including: Acquire a target real-time behavior feature vector of the target real-time behavior feature parameter according to the predetermined driving behavior feature vector quantization strategy; The target real-time behavior feature vector is input into the driving behavior risk model to obtain the initial real-time behavior risk level.

5. The method for predicting student risk based on driving behavior modeling according to claim 4 is characterized in that: Inputting the target real-time behavior feature vector into the driving behavior risk model to obtain the initial real-time behavior risk level includes: Build a driving behavior risk database; Extracting a first data group from the driving behavior risk database, the first data group including a first behavior feature vector with a high risk identifier; Extracting a second data group from the driving behavior risk database, wherein the second data group includes a second behavior feature vector having a medium risk flag; Extracting a third data group from the driving behavior risk database, wherein the third data group includes a third behavior feature vector having a low risk identifier; A training data set is formed according to the first behavior feature vector with a high risk identifier, the second behavior feature vector with a medium risk identifier, and the third behavior feature vector with a low risk identifier; The driving behavior risk model is obtained by performing supervised learning and testing on the training data set based on the support vector machine principle.

6. The method for predicting student risk based on driving behavior modeling according to claim 4 is characterized in that: Inputting the target real-time behavior feature vector into the driving behavior risk model to obtain the initial real-time behavior risk level, and then further comprising: Obtaining a first cosine similarity between the target real-time behavior feature vector and the first historical behavior feature vector; When the first cosine similarity reaches a predetermined similarity threshold, taking the first historical behavior risk level corresponding to the first historical behavior feature vector as the second initial real-time behavior risk level of the target student; The initial real-time behavior risk level is verified with the second initial real-time behavior risk level.

7. The student risk prediction method based on driving behavior modeling according to claim 1 is characterized in that: The predetermined multidimensional risk factor set includes a road risk factor set and an environmental risk factor set, and the road risk factor set includes road flatness, road tortuosity and road slope, and the environmental risk factor set includes wind force, wind speed, temperature and humidity.

8. The student risk prediction method based on driving behavior modeling according to claim 1 is characterized in that: The initial real-time behavior risk level is calibrated according to a predetermined multi-dimensional risk factor set to obtain a target real-time behavior risk level, and a target dynamic risk distribution map of the target student is generated, including: Acquire a target driving area for the target student; Dividing the target driving area to obtain a target area grid set; Collecting risks for a first grid in the target area grid set according to the predetermined multidimensional risk factor set, and performing variance weighting to obtain a first risk index; Using the first risk index as a weight coefficient, adjusting the target real-time behavior risk level to obtain a first predicted behavior risk level of the target student in the first grid; The first predicted behavior risk level is rendered to the first grid, and the target dynamic risk distribution map is generated.

Images