Inspection Method, Device, Computer Equipment and Readable Storage Medium for Fare Evasion Behavior

Abstract

This application relates to a method, device, computer equipment and readable storage medium for inspecting toll evasion behavior. The method includes: obtaining a plurality of preset toll evasion features related to passing through an ETC lane, and determining an initial feature vector of a target vehicle for each preset toll evasion feature; obtaining preset toll evasion weights for each of the plurality of preset toll evasion features; according to the preset toll evasion weights and the initial feature vector of the target vehicle for each preset toll evasion feature, determining a target feature vector of the target vehicle for each preset toll evasion feature; based on the target feature vector of the target vehicle for each preset toll evasion feature, determining the probability that the target vehicle has toll evasion behavior when passing through the ETC lane. Using this method can not only improve the speed and efficiency of identifying toll evading vehicles, but also early warn potential toll evasion behaviors, provide ex-ante inspection means, and reduce the cost and difficulty of ex-post processing.

Description

Technical Field

[0001] The present application relates to the field of computer technologies, and particularly to a method and apparatus for inspecting toll evasion behaviors, a computer device, a computer-readable storage medium, and a computer program product. Background Art

[0002] The ETC system has advantages such as fast speed, safety, environmental protection, and high efficiency. Especially with the implementation of ETC networking across the national expressways, more and more vehicle owners choose to apply for and install ETC cards to use the ETC system to pass through expressway toll stations, which greatly improves people's travel efficiency, especially during holidays, weekends and other periods.

[0003] However, although the ETC non-stop toll collection alleviates traffic problems to a certain extent, some toll-evading customers take advantage of this convenience to evade tolls. The large loss of tolls has attracted great attention from the expressway management department, and strict inspections of passing vehicles have begun. Most of the current inspection methods mainly rely on on-site manual inspections supplemented by back-end inspections. The former not only has low efficiency and few detections, while the latter has a lag and fails to play a good deterrent role. Summary of the Invention

[0004] Based on this, in view of the above technical problems, it is necessary to provide a method and apparatus for inspecting toll evasion behaviors, a computer device, a computer-readable storage medium, and a computer program product that can improve the accuracy and timeliness of high-speed ETC inspections.

[0005] In a first aspect, the present application provides a method for inspecting toll evasion behaviors, including:

[0006] Obtain a plurality of preset toll evasion features related to the passing ETC lane, and determine an initial feature vector of the target vehicle for each of the preset toll evasion features;

[0007] Obtain the preset toll evasion weights of the plurality of preset toll evasion features respectively;

[0008] According to the preset toll evasion weights and the initial feature vectors of the target vehicle for each of the preset toll evasion features, determine the target feature vectors of the target vehicle for each of the preset toll evasion features;

[0009] Based on the target feature vectors of the target vehicle for each of the preset toll evasion features, determine the probability that the target vehicle has a toll evasion behavior when passing through the ETC lane.

[0010] In one of the embodiments, the determining the probability that the target vehicle has a toll evasion behavior when passing through the ETC lane based on the target feature vectors of the target vehicle for each of the preset toll evasion features includes:

[0011] Obtain the preset slope parameter corresponding to each of the preset fare evasion features; for each of the preset fare evasion features, perform a multiplication process on the preset slope parameter corresponding to the targeted preset fare evasion feature and the target feature vector to obtain the to-be-accumulated feature vector corresponding to the targeted preset fare evasion feature; accumulate the to-be-accumulated feature vectors corresponding to each of the preset fare evasion features to obtain an accumulated sum feature vector; obtain a preset intercept parameter, and based on the preset intercept parameter and the accumulated sum feature vector, determine the probability that the target vehicle has a fare evasion behavior when passing through the ETC lane.

[0012] In one embodiment, the obtaining the preset intercept parameter and based on the preset intercept parameter and the accumulated sum feature vector to determine the probability that the target vehicle has a fare evasion behavior when passing through the ETC lane includes:

[0013] Perform an addition process on the preset intercept parameter and the accumulated sum feature vector to obtain a linear feature vector; perform an exponential transformation on the linear feature vector to obtain a non-linear feature vector; based on the non-linear feature vector, determine the probability that the target vehicle has a fare evasion behavior when passing through the ETC lane.

[0014] In one embodiment, obtain a preset probability threshold, and compare the probability that the target vehicle has a fare evasion behavior when passing through the ETC lane with the preset probability threshold; in the case where the probability that the target vehicle has a fare evasion behavior when passing through the ETC lane is greater than or equal to the preset probability threshold, classify the fare evasion behavior of the target vehicle when passing through the ETC lane based on the target feature vector of the target vehicle for each of the preset fare evasion features.

[0015] In one embodiment, the classifying the fare evasion behavior of the target vehicle when passing through the ETC lane based on the target feature vector of the target vehicle for each of the preset fare evasion features includes:

[0016] Obtain a first weight value and a first bias value corresponding to the fare evasion behavior classification category; for each of the preset fare evasion features, based on the first weight value and the first bias value, convert the target feature vector of the targeted preset fare evasion feature into an activation value corresponding to the fare evasion behavior classification category; based on the activation value corresponding to the fare evasion behavior classification category, obtain a fare evasion behavior probability distribution feature vector corresponding to the targeted preset fare evasion feature; classify the fare evasion behavior of the target vehicle when passing through the ETC lane according to the fare evasion behavior probability distribution feature vectors corresponding to each of the preset fare evasion features.

[0017] In one embodiment, the obtaining a plurality of preset fare evasion features related to passing through the ETC lane includes:

[0018] Obtain multiple candidate toll evasion features related to the passing ETC lane; for each of the candidate toll evasion features, output the targeted candidate toll evasion feature to a toll evasion decision tree, where the toll evasion decision tree includes at least one toll evasion node; determine the reduction in impurity generated by the targeted candidate toll evasion feature when passing through each of the toll evasion nodes in the toll evasion decision tree; based on the reduction in impurity generated at each of the toll evasion nodes, determine the average reduction in impurity of the targeted candidate toll evasion feature; based on the average reduction in impurity corresponding to each of the multiple candidate toll evasion features, screen out multiple preset toll evasion features from the multiple candidate toll evasion features.

[0019] In a second aspect, the present application further provides a toll evasion behavior inspection device, including:

[0020] A first determination module, configured to obtain multiple preset toll evasion features related to the passing ETC lane, and determine an initial feature vector of a target vehicle for each of the preset toll evasion features;

[0021] An acquisition module, configured to acquire the preset toll evasion weights of each of the multiple preset toll evasion features;

[0022] A second determination module, configured to determine a target feature vector of the target vehicle for each of the preset toll evasion features according to the preset toll evasion weight and the initial feature vector of the target vehicle for each of the preset toll evasion features;

[0023] A third determination module, configured to determine the probability that the target vehicle has a toll evasion behavior when passing through the ETC lane based on the target feature vector of the target vehicle for each of the preset toll evasion features.

[0024] In a third aspect, the present application further provides a computer device, including a memory and a processor, where the memory stores a computer program, and when the processor executes the computer program, the following steps are implemented:

[0025] Obtain multiple preset toll evasion features related to the passing ETC lane, and determine an initial feature vector of a target vehicle for each of the preset toll evasion features;

[0026] Acquire the preset toll evasion weights of each of the multiple preset toll evasion features;

[0027] Determine a target feature vector of the target vehicle for each of the preset toll evasion features according to the preset toll evasion weight and the initial feature vector of the target vehicle for each of the preset toll evasion features;

[0028] Determine the probability that the target vehicle has a toll evasion behavior when passing through the ETC lane based on the target feature vector of the target vehicle for each of the preset toll evasion features.

[0029] Fourthly, the present application also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the following steps are implemented:

[0030] Obtain a plurality of preset toll evasion features related to passing through an ETC lane, and determine an initial feature vector of the target vehicle for each of the preset toll evasion features;

[0031] Obtain the preset toll evasion weights of the respective preset toll evasion features;

[0032] According to the preset toll evasion weights and the initial feature vectors of the target vehicle for each of the preset toll evasion features, determine the target feature vectors of the target vehicle for each of the preset toll evasion features;

[0033] Based on the target feature vectors of the target vehicle for each of the preset toll evasion features, determine the probability that the target vehicle has a toll evasion behavior when passing through the ETC lane.

[0034] Fifthly, the present application also provides a computer program product, including a computer program. When the computer program is executed by a processor, the following steps are implemented:

[0035] Obtain a plurality of preset toll evasion features related to passing through an ETC lane, and determine an initial feature vector of the target vehicle for each of the preset toll evasion features;

[0036] Obtain the preset toll evasion weights of the respective preset toll evasion features;

[0037] According to the preset toll evasion weights and the initial feature vectors of the target vehicle for each of the preset toll evasion features, determine the target feature vectors of the target vehicle for each of the preset toll evasion features;

[0038] Based on the target feature vectors of the target vehicle for each of the preset toll evasion features, determine the probability that the target vehicle has a toll evasion behavior when passing through the ETC lane.

[0039] For the above toll evasion behavior inspection method, device, computer device, computer-readable storage medium and computer program product, by obtaining a plurality of preset toll evasion features related to passing through an ETC lane and determining the initial feature vectors of the target vehicle for each of the preset toll evasion features, various possible toll evasion behavior patterns can be comprehensively captured. This provides basic data support for subsequent data analysis, making it easier to identify and distinguish normal passing vehicles from potential toll evasion vehicles, thereby improving the pertinence and accuracy of inspection.

[0040] Obtain the preset evasion weights for each of the multiple preset evasion features. Assigning weights to different evasion features can highlight those factors that have a greater impact on evasion behavior, ensuring that more importance is attached to important features during the evaluation process. This helps to improve the accuracy of the prediction model, reduce the false alarm rate, be closer to the actual situation, and enhance its applicability and reliability.

[0041] According to the preset evasion weights and the initial feature vectors of the target vehicle for each of the preset evasion features, determine the target feature vectors of the target vehicle for each of the preset evasion features. The process of applying weights to the initial feature vectors to generate target feature vectors is actually adjusting and optimizing the input data to make it better reflect the actual evasion risk. It enhances the ability to understand evasion behavior, makes the prediction results more accurate, and also increases the adaptability to new situations.

[0042] Based on the target feature vectors of the target vehicle for each of the preset evasion features, determine the probability that the target vehicle has an evasion behavior when passing through the ETC lane. The final step is to use the feature vectors optimized by weighting to calculate the evasion probability, which directly outputs the result of whether the vehicle is likely to evade fees. This method not only improves the speed and efficiency of identifying evading vehicles, but also can pre-warn potential evasion behaviors, provide ex-ante inspection means, and reduce the cost and difficulty of ex-post processing. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the following will briefly introduce the drawings required for use in the description of the embodiments of the present application or related technologies. Obviously, the drawings in the following description are only some embodiments of the present application. For those of ordinary skill in the art, without creative efforts, other related drawings can also be obtained based on these drawings.

[0044] Figure 1 It is an application environment diagram of the evasion behavior inspection method in an embodiment;

[0045] Figure 2 It is a flowchart of the evasion behavior inspection method in an embodiment;

[0046] Figure 3 It is a framework diagram of the evasion behavior inspection method in an embodiment;

[0047] Figure 4 It is a flowchart of the evasion behavior inspection method in another embodiment;

[0048] Figure 5 It is a structural block diagram of the evasion behavior inspection device in an embodiment;

[0049] Figure 6 It is the structural block diagram of the fare evasion behavior inspection device in another embodiment;

[0050] Figure 7 It is the internal structure diagram of a computer device in one embodiment. Specific implementation manners

[0051] In order to make the objectives, technical solutions and advantages of the present application more clear and understandable, the present application will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0052] The fare evasion behavior inspection method provided by the embodiments of the present application can be applied to an application environment as Figure 1 shown. Among them, the terminal 102 communicates with the server 104 through a network. The data storage system can store the data that the server 104 needs to process. The data storage system can be integrated on the server 104, or can be placed in the cloud or other network servers. The terminal 102 is used to generate a fare evasion behavior inspection request and send the fare evasion behavior inspection request to the server 104, so that the server 104 determines the probability that the target vehicle has a fare evasion behavior when passing through the ETC lane based on the target feature vector of the target vehicle for each preset fare evasion feature. Among them, the terminal 102 can be but is not limited to various personal computers, laptop computers, smart phones, tablet computers, Internet of Things devices and portable wearable devices. The Internet of Things devices can be smart speakers, smart TVs, smart air conditioners, smart in-vehicle devices, projection devices, etc. The portable wearable devices can be smart watches, smart bracelets, head-mounted devices, etc. The head-mounted device can be a virtual reality (VR) device, an augmented reality (AR) device, a smart glasses, etc. The server 104 can be an independent physical server, or a server cluster or distributed system composed of multiple physical servers, or a cloud server providing cloud computing services.

[0053] In an exemplary embodiment, as Figure 2 shown, a fare evasion behavior inspection method is provided. Taking the method applied to the Figure 1 server 104 in it as an example, the method includes the following steps 202 to step 208. Among them:

[0054] Step 202, obtain a plurality of preset fare evasion features related to passing through the ETC lane, and determine the initial feature vector of the target vehicle for each preset fare evasion feature.

[0055] Among them, the ETC lane refers to the dedicated lane for the electronic toll collection system, which allows vehicles equipped with an On-Board Unit (OBU) and an ETC card to automatically complete the payment of tolls without stopping when passing through the toll station. Understandably, the ETC lane is usually equipped with special reading and writing devices that can identify the OBU on the vehicle and communicate with it to achieve a fast, safe, and efficient toll collection process.

[0056] The preset toll evasion characteristics refer to the attributes or indicators that can characterize the possible toll evasion behavior of a vehicle. These characteristics may be types such as inconsistent vehicle types, inconsistent vehicle and card, moving OBU device, missed collection of truck OBU, swapping labels in the oncoming direction, malicious following, fraudulently obtaining and using cards, no card at the exit, and evading tolls by making a U-turn.

[0057] The target vehicle refers to a specific vehicle that is being evaluated by the inspection system for possible toll evasion behavior. Understandably, the target vehicle will pass through the ETC lane, and its passing information will be collected as the data source for analysis.

[0058] The initial feature vector refers to, for each individual preset toll evasion characteristic, based on the specific situation of the target vehicle, converting its associated attributes or states into a set of quantized values that are easy to process and analyze.

[0059] Specifically, first, based on historical data analysis and known toll evasion behavior patterns, a series of characteristics that can characterize possible toll evasion behavior are determined. These characteristics can be inconsistent vehicle types, inconsistent vehicle and card, moving OBU device, etc. Ensure that all potential toll evasion behavior types can be captured, so as to comprehensively cover the inspection requirements.

[0060] Then, whenever a vehicle (the target vehicle) passes through the ETC lane, the relevant information of the target vehicle, such as vehicle type, number of axles, import and export times, total weight, etc., is automatically recorded. This information will serve as the basic data for subsequent analysis and provide the necessary raw data support for each preset toll evasion characteristic for quantization processing.

[0061] Finally, for each selected preset toll evasion characteristic, it is converted into one or more quantizable indicators. For example, the vehicle type code is converted into a dummy variable, or continuous values such as the time interval and weight measured directly by the sensor are used. Finally, for each preset toll evasion characteristic, an initial feature vector is constructed according to its corresponding quantization result. This vector contains all the relevant quantized information of the target vehicle on this characteristic.

[0062] In an example, taking a certain preset toll evasion characteristic "inconsistent vehicle types" as an example, the initial feature vector may be [0, 1], indicating that the actual vehicle type of the target vehicle is a small passenger car, while the vehicle type set in the OBU is a truck.

[0063] Step 204: Obtain the preset evasion weights for each of the multiple preset evasion features.

[0064] The preset evasion weight refers to a value assigned to each preset evasion feature, which is used to represent the importance or influence of this feature in determining whether a vehicle has committed an evasion behavior. The purpose is to reflect the different contribution degrees of different features to the identification of evasion behaviors.

[0065] Specifically, first collect a large amount of historical passing records, namely raw data, including known evasion cases and normal passing records. Further clean the raw data, handle missing values and error values, and delete useless variables. Extract effective feature attribute information fields, such as vehicle type, number of axles, inlet and outlet times, total weight, etc. And perform virtualization processing on categorical variables (such as vehicle type) and normalization processing on continuous value variables (such as total weight). It can be understood that the purpose of this step is to ensure the data quality and consistency for weight calculation and improve the effectiveness of model training.

[0066] Then apply the selected weights to train a model (such as a random forest model) and train the processed data. Analyze the feature importance scores output by the weight training model as the basis for the initial weight assignment. Then adjust the parameters through multiple iterations to optimize the model performance until satisfactory feature weights are obtained. These feature weights are used to provide a stable and verified weight configuration for future evasion behavior prediction.

[0067] Step 206: Determine the target feature vectors of the target vehicle for each preset evasion feature according to the preset evasion weights and the initial feature vectors of the target vehicle for each preset evasion feature.

[0068] The target feature vector refers to a quantitative representation that combines the preset evasion weights and is used to more accurately describe the state of the target vehicle for each preset evasion feature. Specifically, it is a new feature vector formed by adjusting the importance of each feature by introducing weights on the basis of the initial feature vector. This process can be regarded as a weighted processing of the original data, aiming to make those features that are more important for identifying evasion behaviors occupy a greater proportion in the final analysis.

[0069] Specifically, traverse each preset evasion feature and its corresponding initial feature vector. For each preset evasion feature, perform the following steps:

[0070] Take out the preset evasion weight corresponding to this preset evasion feature, multiply each element in the initial feature vector by the corresponding weight value to obtain the weighted feature value, and store the weighted feature value in the previously initialized target feature vector container. By processing each feature one by one in this way, ensure that each preset evasion feature is correctly weighted and updated into the target feature vector.

[0071] In one example, assume that the preset fare evasion feature is the vehicle type, and the preset fare evasion weight corresponding to this preset fare evasion feature is 0.3. Then, for a specific target vehicle, its data for this preset fare evasion feature is [minibus]. Further, convert this preset fare evasion feature into a quantifiable index. Assume that this quantifiable index is [1, 0] (binary representation). Then, after combining the weights, the obtained target feature vector is [1×0.3, 1×0.3].

[0072] Step 208: Based on the target feature vectors of the target vehicle for each preset fare evasion feature, determine the probability that the target vehicle has a fare evasion behavior when passing through the ETC lane.

[0073] Among them, the fare evasion behavior refers to the behavior that when a vehicle passes through the ETC (Electronic Toll Collection) lane, it takes certain improper means to avoid the tolls payable. Such behavior violates the normal toll collection rules, resulting in the highway operator failing to receive the due tolls, thus causing economic losses and management troubles.

[0074] Specifically, use a fare evasion prediction model (such as a certain logistic regression model) to predict whether the target vehicle has a fare evasion behavior when passing through the ETC lane. Specifically, for each preset fare evasion feature of the target vehicle, input the target feature vector corresponding to each preset fare evasion feature into the fare evasion prediction model. Through the fare evasion prediction model, process the target feature vector corresponding to each preset fare evasion feature to obtain the probability result output by the fare evasion prediction model. This probability result reflects the probability that the target vehicle has a fare evasion behavior.

[0075] In one example, use a fare evasion prediction model to perform a probability prediction on the fare evasion behavior. For the preset fare evasion feature of the target vehicle, that is, the vehicle type, the target feature vector corresponding to this preset fare evasion feature may be [0, 1]. After inputting this target feature vector into the fare evasion prediction model, the result output by the fare evasion prediction model is 0.75, which means that the target vehicle has a 75% probability of having a fare evasion behavior.

[0076] In one of the embodiments, obtain the preset slope parameter corresponding to each preset fare evasion feature; for each preset fare evasion feature, perform a multiplication process on the preset slope parameter and the target feature vector corresponding to the preset fare evasion feature to obtain the to-be-accumulated feature vector corresponding to the preset fare evasion feature; perform an accumulation process on the to-be-accumulated feature vectors corresponding to each preset fare evasion feature to obtain an accumulated sum feature vector; obtain a preset intercept parameter, and based on the preset intercept parameter and the accumulated sum feature vector, determine the probability that the target vehicle has a fare evasion behavior when passing through the ETC lane.

[0077] Among them, the preset slope parameter refers to the weight or coefficient in the logistic regression model. Each preset fare evasion feature corresponds to a slope parameter, which determines the degree of influence of this feature on the final prediction result.

[0078] The feature vector to be accumulated refers to the result obtained by multiplying a certain preset fare evasion feature by its corresponding preset slope parameter. Understandably, for each preset fare evasion feature, there will be such a vector, which represents the contribution degree of this preset fare evasion feature to the prediction of the fare evasion probability.

[0079] The accumulated sum feature vector refers to the result of adding the results (i.e., each feature vector to be accumulated) obtained by multiplying each preset fare evasion feature by their respective preset slope parameters to obtain a total feature vector. Understandably, the accumulation process is part of calculating the linear combination and is used to generate the linear component of the predicted value in logistic regression.

[0080] The preset intercept parameter refers to the bias term or intercept in the logistic regression model. The preset intercept parameter, as a constant term, does not depend on any input features but is a fixed part of the model and is used to adjust the basic level of the output result.

[0081] Specifically, first, it is necessary to clarify what the preset fare evasion features are. These preset fare evasion features include but are not limited to the total amount, toll, overload rate, vehicle type, total number of axles, etc., and can be specifically set according to actual needs. Obtain the preset slope parameter (i.e., weight or coefficient) corresponding to each preset fare evasion feature from the fare evasion prediction model, as well as a preset intercept parameter. Then, for each preset fare evasion feature, multiply the preset slope parameter corresponding to this preset fare evasion feature by the target feature vector of the target vehicle to generate the feature vector to be accumulated corresponding to this preset fare evasion feature. For example, if there is a preset fare evasion feature of "total amount", then multiply the target feature vector corresponding to the preset fare evasion feature of "total amount" by the preset slope parameter corresponding to the preset fare evasion feature of "total amount". Next, sum up the feature vectors to be accumulated corresponding to all preset fare evasion features to obtain a total accumulated sum feature vector. This step actually constructs a part of the linear prediction function, representing the comprehensive influence of all preset fare evasion features working together. And obtain the preset intercept parameter set in advance, and add this preset intercept parameter to the accumulated sum feature vector. Understandably, this preset intercept parameter represents the basic prediction level of the model even when all features do not exist. Finally, in order to convert the result of the linear combination into a probability form, use the activation function to map the result obtained by adding the preset intercept parameter to the accumulated sum feature vector to the interval (0, 1) to obtain the probability of fare evasion occurring.

[0082] Since the feature vector of the target vehicle is processed by a logistic regression model using a preset slope parameter (weight) and an intercept parameter, the influence of each feature on the toll evasion behavior can be effectively quantified. This method can more accurately capture which features are highly correlated with the toll evasion behavior, thus improving the accuracy of predicting the toll evasion behavior.

[0083] In one embodiment, the preset intercept parameter and the cumulative sum feature vector are added to obtain a linear feature vector; the linear feature vector is subjected to an exponential transformation to obtain a non-linear feature vector; based on the non-linear feature vector, the probability that the target vehicle has a toll evasion behavior when passing through the ETC lane is determined.

[0084] The linear feature vector refers to the result obtained by adding the preset intercept parameter and the cumulative sum feature vector. This process actually constructs the linear part of the logistic regression model, that is, a linear combination, representing the comprehensive linear influence of all preset toll evasion features on whether the target vehicle will have a toll evasion behavior.

[0085] The non-linear feature vector refers to the output after the exponential transformation. The "non-linear feature vector" in this application is actually the result of performing an exponential transformation on the linear feature vector, aiming to map it to a probability space.

[0086] Specifically, first, with the preset intercept parameter and the cumulative sum feature vector as inputs, the preset intercept parameter and the cumulative sum feature vector are added to obtain a linear combination, that is, a linear feature vector. Then, an activation function (Sigmoid Function, S-shaped function) is applied to the obtained linear combination to calculate the probability that the target vehicle has a toll evasion behavior when passing through the ETC lane.

[0087] Optionally, according to the toll evasion prediction model, the calculation formula for the toll evasion probability P(y = 1|x) is:

[0088] P(y = 1|x) = 1 / (1 +

[0089] where A is the preset intercept parameter, B is the preset slope parameter corresponding to each preset toll evasion feature, and xi is the target feature vector corresponding to each preset toll evasion feature.

[0090] Since a linear feature vector is obtained by adding a preset intercept parameter and a cumulative sum feature vector. This step allows each feature to have a different degree of influence on the final prediction result according to its importance (determined by the preset slope parameter). An exponential transformation is performed on the linear feature vector, introducing a non-linear element. This step enables the model to capture more complex, non-linear relationships, thereby improving the model's expressive ability and enabling it to better fit the potential patterns in the data. Based on the non-linear feature vector, a logistic regression or a similar method is used to calculate the probability that the target vehicle has a toll evasion behavior when passing through the ETC lane. The output is a probability value between 0 and 1, and this form of result is easy to understand and interpret.

[0091] In one embodiment, a preset probability threshold is obtained, and the probability that the target vehicle has a toll evasion behavior when passing through the ETC lane is compared with the preset probability threshold; in the case where the probability that the target vehicle has a toll evasion behavior when passing through the ETC lane is greater than or equal to the preset probability threshold, based on the target feature vector of the target vehicle for each preset toll evasion feature, the toll evasion behavior of the target vehicle when passing through the ETC lane is classified.

[0092] The preset probability threshold refers to a critical value preset in the ETC toll inspection method. This critical value is used to distinguish whether a vehicle may be involved in a toll evasion behavior. Specifically, it is a numerical value between 0 and 1, representing the lower limit of the probability that the system determines a vehicle as a toll evasion vehicle.

[0093] Specifically, first, the preset probability threshold is read from the system's configuration file or other set sources. Optionally, the preset probability threshold can be a critical value determined in advance according to experience and data analysis to distinguish whether the target vehicle may be involved in a toll evasion behavior. Then, it is compared whether the probability of the target vehicle's toll evasion behavior calculated in the above step is greater than or equal to the preset probability threshold. Further, if the probability of the target vehicle's toll evasion behavior is greater than or equal to the preset probability threshold, the next step is entered; otherwise, the process ends and the vehicle is considered a normal vehicle. Finally, when the probability of the target vehicle's toll evasion behavior is greater than or equal to the preset probability threshold, the target feature vector is received as input through the toll evasion behavior classification model, and the specific type of toll evasion behavior of the target vehicle is used as the classification result.

[0094] Due to the reasonable setting of the preset probability threshold, it can help reduce the risk of misjudgment, ensuring that neither real toll evaders are let go nor unnecessary interference is caused to normal users. And for those vehicles considered to have a high probability of toll evasion, their types of toll evasion are specifically classified (such as counterfeiting, cheating, defect, and violation types). This helps to deeply understand the toll evasion patterns and formulate more targeted preventive measures.

[0095] In one embodiment, a first weight value and a first bias value corresponding to the classification category of the toll evasion behavior are obtained; for each preset toll evasion feature, based on the first weight value and the first bias value, the target feature vector of the preset toll evasion feature is converted into an activation value corresponding to the classification category of the toll evasion behavior; based on the activation value corresponding to the classification category of the toll evasion behavior, a toll evasion behavior probability distribution feature vector of the preset toll evasion feature is obtained; according to the toll evasion behavior probability distribution feature vectors corresponding to each preset toll evasion feature, the toll evasion behavior of the target vehicle when passing through the ETC lane is classified.

[0096] Among them, the classification category of the toll evasion behavior refers to the classification of different types of toll evasion behaviors that may occur on the ETC lane. For example, counterfeiting, cheating, defect, and violation, etc. Each category represents a specific type of toll evasion behavior pattern. In this application, it is used to further classify the suspected toll evasion vehicles into specific toll evasion types, so as to take more targeted management measures.

[0097] The first weight value refers to the initial weight setting related to the connection between the input layer and the hidden layer or directly to the output layer in the toll evasion behavior classification model. Understandably, the first weight value determines the influence degree of each preset toll evasion feature on the final classification result.

[0098] As another important parameter in the toll evasion behavior classification model, the first bias value refers to the bias term applied to the input layer or the first hidden layer in the toll evasion behavior classification model. It allows the model to activate some neurons even when all inputs are zero. Greatly increases the flexibility of the model, so that even in the absence of input signals, it can produce non-zero outputs, which helps to improve the fitting ability of the model.

[0099] The activation value refers to the result obtained after applying the activation function. Understandably, the activation function is used to introduce non-linear elements, enabling the model to capture more complex patterns. In this application, for each preset toll evasion feature, based on the first weight value and the first bias value, the activation value corresponding to the target feature vector is calculated. This step usually occurs in each layer of the neural network as part of the conversion from input to output.

[0100] The toll evasion behavior probability distribution feature vector refers to a vector representing the probability distribution of the target vehicle belonging to each classification category of the toll evasion behavior. Each element corresponds to a specific classification category of the toll evasion behavior and reflects the likelihood of the target vehicle committing toll evasion under that category.

[0101] Specifically, first, extract or load the pre-trained first weight values and first bias values corresponding to each fare evasion behavior classification category from the already trained fare evasion behavior classification model. It can be understood that these parameters are learned through a large amount of historical data during the model training process and are used to describe the relationship between features and fare evasion types. Each fare evasion behavior classification category corresponds to a weight matrix and a bias vector. Then, for each preset fare evasion feature, use the formula Z = W i ×x i +b i to calculate the linear combination Z, where x i is the target feature vector, and W i and b i are the weight matrix and bias vector corresponding to this feature respectively. Next, apply the activation function to convert the linear combination Z into an activation value a. Each preset fare evasion feature will correspond to an activation value a. Further, use the softmax function () or other suitable normalization methods to convert the activation value a into a vector P representing the probability distribution of each fare evasion behavior classification category (i.e., the fare evasion behavior probability distribution feature vector). It can be understood that this step ensures that the sum of all probabilities is equal to 1, and each element is between 0 and 1, reflecting the probability that the target vehicle belongs to a specific fare evasion type. Finally, comprehensively consider the probability distribution feature vectors P of all preset fare evasion features, and select the fare evasion behavior classification category with the highest probability as the final classification result.

[0102] Since through the comprehensive evaluation of the fare evasion behavior probability distribution feature vector, judgments can be made on multiple feature dimensions, reducing the risk of misjudgment caused by a single feature and improving the reliability of the decision-making. Through this step, not only can it be identified whether a fare evasion behavior occurs, but it can also be further broken down into specific fare evasion types (such as counterfeiting, cheating, defect, and violation types). This comprehensive classification ability helps to deeply understand the fare evasion patterns and provides support for formulating targeted prevention measures.

[0103] In one embodiment, obtain multiple candidate fare evasion features related to the ETC lane passing; for each candidate fare evasion feature, output the candidate fare evasion feature to the fare evasion decision tree, and the fare evasion decision tree includes at least one fare evasion node; determine the impurity reduction amount generated by the split of the candidate fare evasion feature when passing through each fare evasion node in the fare evasion decision tree; based on the impurity reduction amount generated by the split at each fare evasion node, determine the average impurity reduction amount of the candidate fare evasion feature; based on the average impurity reduction amounts corresponding to the multiple candidate fare evasion features, screen out multiple preset fare evasion features from the multiple candidate fare evasion features.

[0104] Among them, candidate toll-evasion features refer to various feature variables that may be related to toll-evasion behaviors on ETC lanes. These features can include, but are not limited to, vehicle type, total weight, toll amount, overloading rate, import and export times, etc.

[0105] The toll-evasion decision tree refers to a machine learning model that divides the dataset through a series of conditional judgments (nodes) and finally assigns samples to different categories. In this application, the toll-evasion decision tree is specifically used to analyze and classify toll-evasion behaviors.

[0106] A toll-evasion node refers to an internal node in the toll-evasion decision tree. Each node corresponds to a specific candidate toll-evasion feature and divides the data according to the value of this feature. In this application, each toll-evasion node divides the data entering the node into two subsets by applying a condition (such as a threshold of a certain feature), and then recursively constructs a finer-grained subtree until reaching the leaf node.

[0107] The impurity reduction amount refers to a measure of how much the "disorder degree" (impurity) of the data distribution is reduced after splitting using a certain candidate toll-evasion feature at a certain node of the toll-evasion decision tree.

[0108] The average impurity reduction amount refers to the average of the impurity reduction amounts of a certain candidate toll-evasion feature at all relevant nodes in the entire toll-evasion decision tree. It is one of the key indicators for evaluating the importance of this feature. By calculating the average impurity reduction amount for each candidate toll-evasion feature, the influence of each feature on the classification effect can be quantified, helping to screen out the most predictive preset toll-evasion features.

[0109] Specifically, first, all features in the ETC system that may be related to toll-evasion behaviors are used as candidate toll-evasion features. These candidate toll-evasion features can include, but are not limited to, vehicle type, total weight, toll amount, overloading rate, import and export times, etc.

[0110] Then, for each candidate toll-evasion feature, construct or use an existing toll-evasion decision tree model. This model includes at least one toll-evasion node (i.e., the root node) for evaluating the effect of a specific feature in distinguishing toll-evasion behaviors. And each toll-evasion node corresponds to a conditional test to determine how the data is divided into different child nodes. It can be understood that the performance of each candidate toll-evasion feature in the toll-evasion decision tree, especially how it affects the data distribution.

[0111] Next, for each candidate toll-evasion feature, calculate the change in impurity before and after splitting at each toll-evasion node of the toll-evasion decision tree. The specific calculation formula is as follows:

[0112]

[0113] Among them, I(N) is the original impurity of node N, P is the data ratio that reaches each child node N after splitting and I(N ) is the reduction in impurity of child node N .

[0114] Finally, for each candidate fare evasion feature, calculate the average of the reduction in impurity on all relevant fare evasion nodes. This step aims to quantify the importance of each feature in the entire decision tree. Further, sort all candidate fare evasion features according to the magnitude of the average reduction in impurity. Select several features with the highest average reduction in impurity as the final preset fare evasion features. These features are considered to be the key features that can most effectively distinguish fare evasion behaviors.

[0115] By calculating the reduction in impurity generated during the splitting of each candidate fare evasion feature at each node in the fare evasion decision tree and further calculating the average reduction in impurity, it is possible to quantify the impact of each feature on the classification result. This method is based on data and statistical analysis, ensuring the scientificity and objectivity of the feature selection process. And using the decision tree model for feature evaluation, its structure is clear and easy to interpret. The conditional tests at each node intuitively show how the features affect the classification result, enabling both managers and technicians to understand the working principle of the model. Finally, by screening out the most important preset fare evasion features, the interference of irrelevant or redundant features on model training is reduced, thereby improving the overall performance of the classification model, including indicators such as accuracy and recall.

[0116] In one embodiment, referring to Figure 3 , the processing flow of the ETC toll inspection system mainly includes the following stages:

[0117] Data collection and analysis stage: First, start with the data stored in the highway network toll collection system. Conduct preliminary collection and analysis of this data to identify the classification of fare evasion behaviors.

[0118] Abnormal vehicle feature analysis stage: On the basis of data collection, further analyze the features of abnormal vehicles. This includes multi-dimensional information such as overall features, time features, spatial features, and toll features.

[0119] Data preprocessing stage: A key step to ensure the accuracy of subsequent model training through data preprocessing. It includes data normalization (making data from different sources comparable), data discretization (converting continuous variables into categorical variables), and dataset sampling (selecting representative samples for modeling).

[0120] Determination of the fare evasion vehicle dataset stage: Through the above steps, finally determine the dataset for training and testing the fare evasion vehicle recognition model.

[0121] The stage of the toll evasion vehicle data training set: Use the random forest algorithm for feature selection to find the key factors affecting toll evasion. Based on the selected features, use the logistic regression model to predict the toll evasion probability. Finally, use the neural network model to classify the types of toll evasion.

[0122] The model verification and analysis stage: Verify the established model, evaluate its performance in actual applications, and make necessary adjustments and optimizations.

[0123] The stage of toll evasion vehicle identification: Apply the trained and verified model to the actual scenario to achieve effective identification of toll evasion vehicles.

[0124] The stage of the highway ETC business system: Finally, the identification results are fed back to the highway ETC business system to help the management department take timely measures to deal with toll evasion behavior and maintain normal traffic order.

[0125] Through the entire process from the original data to the final application in the above steps, the important role of using big data analysis technology in improving the efficiency of ETC toll inspection is realized.

[0126] In an exemplary embodiment, as Figure 4 shown, it includes steps 402 to 410. Among them:

[0127] Step 402, obtain a plurality of candidate toll evasion features related to the passing ETC lane; for each candidate toll evasion feature, output the candidate toll evasion feature to the toll evasion decision tree, and the toll evasion decision tree includes at least one toll evasion node; determine the reduction in impurity generated by the split of the candidate toll evasion feature when passing through each toll evasion node in the toll evasion decision tree; based on the reduction in impurity generated by the split at each toll evasion node, determine the average reduction in impurity of the candidate toll evasion feature; based on the average reduction in impurity corresponding to each of the plurality of candidate toll evasion features, screen out a plurality of preset toll evasion features from the plurality of candidate toll evasion features; determine the initial feature vector of the target vehicle for each preset toll evasion feature;

[0128] Step 404, obtain the preset toll evasion weights of each of the plurality of preset toll evasion features;

[0129] Step 406, according to the preset toll evasion weights and the initial feature vectors of the target vehicle for each preset toll evasion feature, determine the target feature vectors of the target vehicle for each preset toll evasion feature;

[0130] Step 408: Obtain the preset slope parameter corresponding to each preset fare evasion feature; for each preset fare evasion feature, perform a multiplication process on the preset slope parameter corresponding to the targeted preset fare evasion feature and the target feature vector to obtain the to-be-accumulated feature vector corresponding to the targeted preset fare evasion feature; perform an accumulation process on the to-be-accumulated feature vectors corresponding to each preset fare evasion feature to obtain the accumulated sum feature vector; obtain the preset intercept parameter, and perform an addition process on the preset intercept parameter and the accumulated sum feature vector to obtain the linear feature vector; perform an exponential transformation on the linear feature vector to obtain the non-linear feature vector; based on the non-linear feature vector, determine the probability that the target vehicle has a fare evasion behavior when passing through the ETC lane.

[0131] Step 410: Obtain the preset probability threshold, and compare the probability that the target vehicle has a fare evasion behavior when passing through the ETC lane with the preset probability threshold; in the case where the probability that the target vehicle has a fare evasion behavior when passing through the ETC lane is greater than or equal to the preset probability threshold, obtain the first weight value and the first bias value corresponding to the fare evasion behavior classification category; for each preset fare evasion feature, based on the first weight value and the first bias value, convert the target feature vector of the targeted preset fare evasion feature into the activation value corresponding to the fare evasion behavior classification category; based on the activation value corresponding to the fare evasion behavior classification category, obtain the fare evasion behavior probability distribution feature vector corresponding to the targeted preset fare evasion feature; classify the fare evasion behavior of the target vehicle when passing through the ETC lane according to the fare evasion behavior probability distribution feature vectors corresponding to each preset fare evasion feature.

[0132] It should be understood that although the steps in the flowcharts involved in the above embodiments are shown in sequence according to the arrows, these steps do not necessarily need to be executed in the order indicated by the arrows. Unless there is a clear indication in this article, the execution of these steps does not have a strict order limit, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above embodiments may include multiple steps or multiple stages. These steps or stages do not necessarily need to be executed at the same moment, but can be executed at different moments. The execution order of these steps or stages does not necessarily need to be sequential, but can be executed alternately or in turn with at least a part of other steps or steps or stages in other steps.

[0133] Based on the same inventive concept, an embodiment of the present application also provides a fare evasion behavior inspection device for implementing the above-mentioned fare evasion behavior inspection method. The solution provided by this device for solving problems is similar to the solution described in the above method. Therefore, the specific limitations in one or more embodiments of the following fare evasion behavior inspection device can refer to the limitations on the fare evasion behavior inspection method in the above text, and will not be repeated here.

[0134] In an exemplary embodiment, asFigure 5 As shown in the figure, a toll evasion behavior inspection device 500 is provided, including: a first determination module 502, an acquisition module 504, a second determination module 506, and a third determination module 508, where:

[0135] The first determination module 502 is configured to obtain a plurality of preset toll evasion features related to passing through the ETC lane, and determine an initial feature vector of the target vehicle for each preset toll evasion feature;

[0136] The acquisition module 504 is configured to obtain preset toll evasion weights for each of the plurality of preset toll evasion features;

[0137] The second determination module 506 is configured to determine a target feature vector of the target vehicle for each preset toll evasion feature according to the preset toll evasion weight and the initial feature vector of the target vehicle for each preset toll evasion feature;

[0138] The third determination module 508 is configured to determine the probability that the target vehicle has a toll evasion behavior when passing through the ETC lane based on the target feature vector of the target vehicle for each preset toll evasion feature.

[0139] In one embodiment, the third determination module 508 is configured to obtain a preset slope parameter corresponding to each preset toll evasion feature; for each preset toll evasion feature, perform a multiplication process on the preset slope parameter corresponding to the preset toll evasion feature and the target feature vector to obtain a to-be-accumulated feature vector corresponding to the preset toll evasion feature; perform an accumulation process on the to-be-accumulated feature vectors corresponding to each preset toll evasion feature to obtain an accumulated sum feature vector; obtain a preset intercept parameter, and determine the probability that the target vehicle has a toll evasion behavior when passing through the ETC lane based on the preset intercept parameter and the accumulated sum feature vector.

[0140] In one embodiment, the third determination module 508 is configured to perform an addition process on the preset intercept parameter and the accumulated sum feature vector to obtain a linear feature vector; perform an exponential transformation on the linear feature vector to obtain a non-linear feature vector; determine the probability that the target vehicle has a toll evasion behavior when passing through the ETC lane based on the non-linear feature vector.

[0141] In one embodiment, the toll evasion behavior inspection device further includes a classification module 510, configured to obtain a preset probability threshold, and compare the probability that the target vehicle has a toll evasion behavior when passing through the ETC lane with the preset probability threshold; in the case where the probability that the target vehicle has a toll evasion behavior when passing through the ETC lane is greater than or equal to the preset probability threshold, classify the toll evasion behavior of the target vehicle when passing through the ETC lane based on the target feature vector of the target vehicle for each preset toll evasion feature.

[0142] In one embodiment, the classification module 510 is configured to obtain a first weight value and a first bias value corresponding to the classification category of the fare evasion behavior; for each preset fare evasion feature, based on the first weight value and the first bias value, convert the target feature vector of the preset fare evasion feature into an activation value corresponding to the fare evasion behavior classification category; based on the activation value corresponding to the fare evasion behavior classification category, obtain a fare evasion behavior probability distribution feature vector of the preset fare evasion feature; and classify the fare evasion behavior of the target vehicle when passing through the ETC lane according to the fare evasion behavior probability distribution feature vectors corresponding to each preset fare evasion feature.

[0143] In one embodiment, the first determination module 502 is configured to obtain a plurality of candidate fare evasion features related to passing through the ETC lane; for each candidate fare evasion feature, output the candidate fare evasion feature to a fare evasion decision tree, where the fare evasion decision tree includes at least one fare evasion node; determine the reduction in impurity generated by splitting when the candidate fare evasion feature passes through each fare evasion node in the fare evasion decision tree; based on the reduction in impurity generated by splitting at each fare evasion node, determine the average reduction in impurity of the candidate fare evasion feature; and screen out a plurality of preset fare evasion features from the plurality of candidate fare evasion features based on the average reduction in impurity corresponding to each of the plurality of candidate fare evasion features.

[0144] In another embodiment, as Figure 6 shown, Figure 6 FIG. is a structural block diagram of a fare evasion behavior inspection device in another embodiment, including: a first determination module 502, an acquisition module 504, a second determination module 506, a third determination module 508, and a classification module 510, where: the classification module 510 is configured to obtain a first weight value and a first bias value corresponding to the classification category of the fare evasion behavior; for each preset fare evasion feature, based on the first weight value and the first bias value, convert the target feature vector of the preset fare evasion feature into an activation value corresponding to the fare evasion behavior classification category; based on the activation value corresponding to the fare evasion behavior classification category, obtain a fare evasion behavior probability distribution feature vector of the preset fare evasion feature; classify the fare evasion behavior of the target vehicle when passing through the ETC lane according to the fare evasion behavior probability distribution feature vectors corresponding to each preset fare evasion feature, and obtain a plurality of candidate fare evasion features related to passing through the ETC lane; for each candidate fare evasion feature, output the candidate fare evasion feature to a fare evasion decision tree, where the fare evasion decision tree includes at least one fare evasion node; determine the reduction in impurity generated by splitting when the candidate fare evasion feature passes through each fare evasion node in the fare evasion decision tree; based on the reduction in impurity generated by splitting at each fare evasion node, determine the average reduction in impurity of the candidate fare evasion feature; and screen out a plurality of preset fare evasion features from the plurality of candidate fare evasion features based on the average reduction in impurity corresponding to each of the plurality of candidate fare evasion features.

[0145] Each module in the above-mentioned fare evasion behavior inspection device can be implemented in whole or in part by software, hardware, or a combination thereof. Each of the above modules can be embedded in the processor of a computer device in hardware form or independent of it, or stored in the memory of the computer device in software form, so that the processor can call and execute the operations corresponding to each of the above modules.

[0146] In an exemplary embodiment, a computer device is provided. The computer device can be a server, and its internal structure diagram can be as Figure 7 shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, abbreviated as I / O), and a communication interface. Among them, the processor, the memory, and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store data related to fare evasion behavior inspection. The input / output interface of the computer device is used to exchange information between the processor and external devices. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, it implements a method for inspecting fare evasion behavior.

[0147] Those skilled in the art can understand that Figure 7 the structure shown in

[0148] is only a block diagram of a part of the structure related to the solution of this application, and does not constitute a limitation on the computer device to which the solution of this application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine some components, or have different component arrangements.

[0149] In an embodiment, a computer device is further provided, including a memory and a processor. A computer program is stored in the memory, and when the processor executes the computer program, it implements the steps in the above method embodiments.

[0150] In an embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by the processor, it implements the steps in the above method embodiments.

[0151] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use, and processing of relevant data need to comply with relevant regulations.

[0152] Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be completed by instructing relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above methods. Among them, any reference to a memory, database, or other medium used in the embodiments provided in this application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc. The databases involved in the embodiments provided in this application can include at least one of relational databases and non-relational databases. Non-relational databases can include distributed databases based on blockchain, etc., without limitation. The processors involved in the embodiments provided in this application can be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, data processing logics based on quantum computing, artificial intelligence (AI) processors, etc., without limitation.

[0153] The technical features of the above embodiments can be combined arbitrarily. For the sake of concise description, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, it should be considered as the scope recorded in this application.

[0154] The above-described embodiments merely represent several implementation manners of this application. The description thereof is relatively specific and detailed, but it should not be construed as a limitation on the scope of the patent of this application. It should be noted that for those of ordinary skill in the art, without departing from the concept of this application, several modifications and improvements can still be made, and these all belong to the protection scope of this application. Therefore, the protection scope of this application shall be subject to the appended claims.

Claims

1. A method for inspecting fare evasion behavior, characterized in that, The method includes: Obtaining a plurality of candidate toll-evasion features related to the ETC lane for passing; for each of the candidate toll-evasion features, outputting the candidate toll-evasion feature targeted to a toll-evasion decision tree, the toll-evasion decision tree including at least one toll-evasion node; calculating the reduction in impurity generated by splitting when the candidate toll-evasion feature targeted passes through each of the toll-evasion nodes in the toll-evasion decision tree, and the specific calculation formula is as follows: Where I(N) is the original impurity of node N, Pi is the data proportion reaching each sub-node Ni after splitting, and I(Ni) is the reduction in impurity of sub-node Ni; Based on the reduction in impurity generated by splitting at each of the toll-evasion nodes, determining the average reduction in impurity of the candidate toll-evasion feature targeted; based on the average reduction in impurity corresponding to each of the plurality of candidate toll-evasion features, screening out a plurality of preset toll-evasion features from the plurality of candidate toll-evasion features, and determining the initial feature vector of the target vehicle for each of the preset toll-evasion features; Obtaining the preset toll-evasion weights corresponding to each of the plurality of preset toll-evasion features; According to the preset toll-evasion weights and the initial feature vector of the target vehicle for each of the preset toll-evasion features, determining the target feature vector of the target vehicle for each of the preset toll-evasion features; Obtaining the preset slope parameter corresponding to each of the preset toll-evasion features, the preset slope parameter characterizing the influence degree of the preset toll-evasion feature on the prediction probability; for each of the preset toll-evasion features, performing a multiplication process on the preset slope parameter and the target feature vector of the preset toll-evasion feature targeted to obtain the to-be-accumulated feature vector corresponding to the preset toll-evasion feature targeted, the to-be-accumulated feature vector characterizing the contribution degree of the preset toll-evasion feature to the toll-evasion probability prediction; performing an accumulation process on the to-be-accumulated feature vectors corresponding to each of the preset toll-evasion features to obtain an accumulated sum feature vector; obtaining a preset intercept parameter, the preset intercept parameter being used to adjust the basic level of the toll-evasion probability, and based on the preset intercept parameter and the accumulated sum feature vector, determining the probability that the target vehicle has a toll-evasion behavior when passing through the ETC lane; Obtaining a first weight value and a first bias value corresponding to the toll-evasion behavior classification category; for each of the preset toll-evasion features, based on the first weight value and the first bias value, converting the target feature vector of the preset toll-evasion feature targeted into an activation value corresponding to the toll-evasion behavior classification category; based on the activation value corresponding to the toll-evasion behavior classification category, obtaining the toll-evasion behavior probability distribution feature vector of the preset toll-evasion feature targeted; classifying the toll-evasion behavior of the target vehicle when passing through the ETC lane according to the toll-evasion behavior probability distribution feature vectors corresponding to each of the preset toll-evasion features.

2. The method according to claim 1, wherein The determining the probability that the target vehicle has a toll-evasion behavior when passing through the ETC lane based on the preset intercept parameter and the accumulated sum feature vector includes: Performing an addition process on the preset intercept parameter and the accumulated sum feature vector to obtain a linear feature vector; Performing an exponential transformation on the linear feature vector to obtain a non-linear feature vector; Based on the non-linear feature vector, determine the probability that the target vehicle has a behavior of evading tolls when passing through the ETC lane.

3. The method according to claim 1, wherein The method further includes: Obtain a preset probability threshold, and compare the probability that the target vehicle has a behavior of evading tolls when passing through the ETC lane with the preset probability threshold; When the probability that the target vehicle has a behavior of evading tolls when passing through the ETC lane is greater than or equal to the preset probability threshold, classify the behavior of the target vehicle evading tolls when passing through the ETC lane based on the target feature vector of the target vehicle for each preset toll-evading feature.

4. The method according to claim 3, characterized in that, The preset probability threshold is a preset critical value used to distinguish whether the target vehicle may be involved in a behavior of evading tolls.

5. A toll evasion behavior inspection device, characterized in that The device includes: A first determination module, configured to obtain a plurality of candidate toll-evading features related to passing through the ETC lane; for each of the candidate toll-evading features, output the candidate toll-evading feature to a toll-evading decision tree, where the toll-evading decision tree includes at least one toll-evading node; calculate the reduction in impurity generated by splitting when the candidate toll-evading feature passes through each toll-evading node in the toll-evading decision tree. The specific calculation formula is as follows: Where I(N) is the original impurity of node N, Pi is the proportion of data reaching each child node Ni after splitting, and I(Ni) is the reduction in impurity of child node Ni; Based on the reduction in impurity generated by splitting at each toll-evading node, determine the average reduction in impurity of the candidate toll-evading feature; based on the average reduction in impurity corresponding to each of the plurality of candidate toll-evading features, screen out a plurality of preset toll-evading features from the plurality of candidate toll-evading features, and determine the initial feature vector of the target vehicle for each preset toll-evading feature; An acquisition module, configured to acquire the preset toll-evading weights corresponding to the plurality of preset toll-evading features; A second determination module, configured to determine the target feature vector of the target vehicle for each preset toll-evading feature according to the preset toll-evading weight and the initial feature vector of the target vehicle for each preset toll-evading feature; A third determination module, configured to acquire a preset slope parameter corresponding to each preset toll-evading feature, where the preset slope parameter characterizes the influence degree of the preset toll-evading feature on the predicted probability; for each preset toll-evading feature, perform a multiplication process on the preset slope parameter of the preset toll-evading feature and the target feature vector to obtain a to-be-accumulated feature vector corresponding to the preset toll-evading feature, where the to-be-accumulated feature vector characterizes the contribution degree of the preset toll-evading feature to the prediction of the toll-evading probability; perform an accumulation process on the to-be-accumulated feature vectors corresponding to each preset toll-evading feature to obtain an accumulated sum feature vector; acquire a preset intercept parameter, where the preset intercept parameter is used to adjust the basic level of the toll-evading probability, and based on the preset intercept parameter and the accumulated sum feature vector, determine the probability that the target vehicle has a behavior of evading tolls when passing through the ETC lane; A classification module, configured to obtain a first weight value and a first bias value corresponding to a toll evasion behavior classification category; for each of the preset toll evasion features, based on the first weight value and the first bias value, convert a target feature vector of the preset toll evasion feature to an activation value corresponding to the toll evasion behavior classification category; based on the activation value corresponding to the toll evasion behavior classification category, obtain a toll evasion behavior probability distribution feature vector of the preset toll evasion feature; classify the toll evasion behavior of the target vehicle when passing through the ETC lane according to the toll evasion behavior probability distribution feature vector corresponding to each of the preset toll evasion features.

6. The device according to claim 5, characterized in that The third determination module is configured to perform an addition process on the preset intercept parameter and the cumulative sum feature vector to obtain a linear feature vector; perform an exponential transformation on the linear feature vector to obtain a non-linear feature vector; based on the non-linear feature vector, determine the probability that the target vehicle has a toll evasion behavior when passing through the ETC lane.

7. The device according to claim 5, characterized in that, The classification module is configured to obtain a preset probability threshold, and compare the probability that the target vehicle has a toll evasion behavior when passing through the ETC lane with the preset probability threshold; in the case that the probability that the target vehicle has a toll evasion behavior when passing through the ETC lane is greater than or equal to the preset probability threshold, classify the toll evasion behavior of the target vehicle when passing through the ETC lane based on the target feature vector of the target vehicle for each of the preset toll evasion features.

8. A computer device, comprising a memory and a processor, the memory storing a computer program, characterized in that, When the processor executes the computer program, the steps of the method according to any one of claims 1 to 4 are implemented.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, the steps of the method according to any one of claims 1 to 4 are implemented.

10. A computer program product comprising a computer program, characterized in that, When the computer program is executed by the processor, the steps of the method according to any one of claims 1 to 4 are implemented.

Images