Traffic risk factor assessment method and device, electronic equipment and storage medium

Abstract

The invention discloses a traffic risk factor assessment method and device, electronic equipment and a storage medium, and the method comprises the steps: obtaining initial data from historical traffic data, the initial data comprising the feature data of an accident road and the type of an actual risk factor causing a traffic risk; performing abnormal data screening and normalization processing on the initial data to obtain sample data; constructing a feature learning model by using a deep learning algorithm; training the feature learning model by taking the sample data as input to obtain a risk assessment model; a deep learning algorithm is adopted to construct a feature learning model, and a machine learning model capable of identifying different risk features is trained in combination with historical sample data, so that traffic risk factor categories of different intersection road sections can be quickly and accurately evaluated and analyzed, and the development of accident risk prevention work of actual roads can be pertinently guided. And traffic risks can be prevented and controlled in time.

Description

Technical Field

[0001] The present invention relates to the technical field of traffic safety maintenance, and particularly to a traffic risk factor assessment method, device, electronic device and storage medium. Background Art

[0002] The rapid development of digital technologies such as big data, cloud computing, and artificial intelligence has brought new favorable opportunities for promoting the solution of technical bottlenecks in previous accident prevention and getting rid of traditional management modes. Among multi-source data such as traffic flow states, driving behaviors, weather conditions, and road environments before an accident, there are often a large amount of traffic risk information, which has important mining value for accident prevention.

[0003] Currently, there are mainly two categories for mining road traffic accident risk factors. One category is the traditional qualitative or qualitative-quantitative combined analysis methods, such as expert experience method, fault tree analysis method, cause-and-effect diagram analysis method, etc.; the other category is data mining and machine learning methods based on more abundant and comprehensive traffic information data to identify and analyze traffic safety risks, such as text data mining, etc.

[0004] However, whether it is the traditional qualitative-quantitative combined analysis method or the analysis method based on data mining of accident investigation reports, it generally focuses on the analysis and evaluation of risk factors after an accident, and it is difficult to predict accident risk factors before the risk occurs, which is not conducive to timely prevention and control of risk hazards. Summary of the Invention

[0005] To solve the above technical problems, the present invention provides a traffic risk factor assessment method.

[0006] In a first aspect, the present invention provides a traffic risk factor assessment method, including:

[0007] Obtain initial data from historical traffic data, where the initial data includes characteristic data of accident roads and actual risk factor categories causing traffic risks;

[0008] Perform abnormal data screening and normalization processing on the initial data to obtain sample data;

[0009] Build a feature learning model using a deep learning algorithm;

[0010] Use the sample data as input to train the feature learning model to obtain a risk assessment model;

[0011] Use the characteristic data of the real-time road as input, and adopt the risk assessment model to identify the traffic risk factor categories of the road.

[0012] In a second aspect, the present invention provides a traffic risk factor assessment device, including:

[0013] An initial data acquisition module, configured to acquire initial data from historical traffic data, where the initial data includes characteristic data of accident roads and actual risk factor categories that cause traffic risks;

[0014] A data processing module, configured to perform abnormal data screening and normalization processing on the initial data to obtain sample data;

[0015] A model construction module, configured to construct a feature learning model using a deep learning algorithm;

[0016] A model training module, configured to use the sample data as input to train the feature learning model to obtain a risk assessment model;

[0017] A model application module, configured to use the characteristic data of real-time roads as input and adopt the risk assessment model to identify the traffic risk factor categories of roads.

[0018] In a third aspect, the present invention provides an electronic device, where the electronic device includes:

[0019] At least one processor; and

[0020] A memory communicatively connected to the at least one processor; wherein,

[0021] The memory stores a computer program executable by the at least one processor, and when the computer program is executed by the at least one processor, the at least one processor is enabled to execute the traffic risk factor assessment method described in the first aspect of the present invention.

[0022] In a fourth aspect, the present invention provides a computer-readable storage medium, where the computer-readable storage medium stores computer instructions, and when the computer instructions are executed by a processor, the traffic risk factor assessment method described in the first aspect of the present invention is implemented.

[0023] A traffic risk factor assessment method provided by an embodiment of the present invention acquires initial data from historical traffic data, where the initial data includes characteristic data of accident roads and actual risk factor categories that cause traffic risks; performs abnormal data screening and normalization processing on the initial data to obtain sample data; constructs a feature learning model using a deep learning algorithm; uses the sample data as input to train the feature learning model to obtain a risk assessment model; constructs a feature learning model using a deep learning algorithm, and trains a machine learning model capable of identifying different risk characteristics in combination with historical sample data, so as to quickly and accurately evaluate and analyze the traffic risk factor categories of different intersection sections, and can specifically guide the development of accident risk prevention work on actual roads, and timely prevent and control traffic risks.

[0024] It should be understood that the content described in this section is not intended to identify the key or important features of the embodiments of the present invention, nor is it used to limit the scope of the present invention. Other features of the present invention will become readily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following will briefly introduce the accompanying drawings required for the description of the embodiments. Obviously, the accompanying drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other accompanying drawings can be obtained based on these drawings.

[0026] Figure 1 is a flowchart of a traffic risk factor assessment method provided in Embodiment 1 of the present invention;

[0027] Figure 2 is a schematic diagram of the network structure of a feature learning model provided in Embodiment 1 of the present invention;

[0028] Figure 3 is a flowchart of a traffic risk factor assessment method provided in Embodiment 2 of the present invention;

[0029] Figure 4 is a schematic diagram of the structure of a traffic risk factor assessment device provided in Embodiment 3 of the present invention;

[0030] Figure 5 is a schematic diagram of the structure of an electronic device provided in Embodiment 4 of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0031] In order to enable those skilled in the art to better understand the solution of the present invention, the following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0032] Embodiment 1

[0033] Figure 1 is a flowchart of a traffic risk factor assessment method provided in Embodiment 1 of the present invention. This embodiment is applicable to the situation of traffic risk factor assessment. This method can be executed by a traffic risk factor assessment device, which can be implemented in the form of hardware and / or software, and the traffic risk factor assessment device can be configured in an electronic device. As Figure 1 shown, the traffic risk factor assessment method includes:

[0034] S101. Obtain initial data from historical traffic data. The initial data includes characteristic data of accident roads and actual risk factor categories that cause traffic risks.

[0035] The characteristic data mainly includes traffic operation information, traffic accident violation information, road physical property information, meteorological information, etc.

[0036] Among them, (1) Traffic operation information: vehicle flow, average vehicle speed, vehicle density, proportion of long-distance passenger and tourist passenger vehicle flow, proportion of heavy freight vehicle flow, proportion of hazardous chemical vehicle flow, proportion of passing vehicle flow, etc.;

[0037] (2) Traffic accident violation information: number of fatal accidents in the past 1 month, number of accidents in the past 12 months, accident weather types in the past 12 months, vehicle usage nature of accidents in the past 12 months, accident forms in the past 12 months, accident determination reasons in the past 12 months, accident time in the past 12 months, accident road surface conditions in the past 12 months, accident road isolation conditions in the past 12 months, number of violations in the past 12 months, violation types in the past 12 months, number of police situations in the past 12 months, etc.;

[0038] (3) Road physical characteristic information: road grade, section length, section width, number of lanes in the section, section curvature, proximity to water, etc.;

[0039] (4) Meteorological information: rainfall amount in the section, snowfall amount, visibility in foggy weather, etc.;

[0040] The actual risk factor categories that cause traffic risks are the risk factor categories given after accident analysis and on-site exploration.

[0041] S102. Perform abnormal data screening and normalization processing on the initial data to obtain sample data.

[0042] Specifically, performing abnormal data screening and normalization processing on the initial data to obtain sample data includes: using the method of drawing box plots to detect outliers in the initial data and deleting the outliers; for non-numerical data in the initial data, converting the non-numerical data into numerical data according to the data type of the non-numerical data; when all the initial data is numerical data, classifying the initial data; and performing normalization processing on each class of the initial data to obtain sample data.

[0043] Data points outside the edge lines in a box plot and data points outside the edge lines in a line plot are called outliers. In a box plot, the edge lines usually consist of the upper edge and the lower edge, and these edge lines represent the minimum and maximum values of the data. Data points located outside these edge lines are considered outliers, which may indicate errors in data collection or processing, or real anomalies, and it is necessary to avoid the adverse effects of outliers on the training effect of the model.

[0044] For different data types, there is a situation of inconsistent dimensions. Normalization processing can eliminate the influence of different feature dimensions, enabling comparison and calculation between different features. In machine learning, normalization processing can accelerate the convergence speed of the model and improve the performance of the model. Normalization processing can map the data within the range of 0 to 1, simplify the calculation process, and improve the data processing efficiency.

[0045] In order to unify the data scale and improve the model recognition effect, in this embodiment, for non-numerical data, including sample accident weather types, accident vehicle usage natures, accident forms, accident determination reasons, accident times, accident road conditions, accident road isolation conditions, illegal types, and sample waterside conditions, etc., they are classified by category and assigned to specific numerical categories; for numerical data, including sample accidents, violations, police situations, traffic flows, densities, speeds, proportions of various vehicle types, etc., the Z-score normalization method is used for processing. This method converts the data into a standard normal distribution with a mean of 0 and a variance of 1 by subtracting the mean and dividing by the standard deviation.

[0046] The conversion formula of the Z-score normalization method is:

[0047] x′ = (x - mean) / std

[0048] Among them, x′ is the normalized data, x is the original data, mean is the mean of the original data, and std is the standard deviation of the original data.

[0049] S103. Construct a feature learning model using a deep learning algorithm.

[0050] Deep learning automatically learns higher-level feature representations from raw data by using deep neural network models. Specifically, the feature learning model includes N convolutional layers, N pooling layers, 1 fully connected layer, and 1 output layer. During the training process, the output end of each convolutional layer is connected to the input end of 1 pooling layer, the output end of the last pooling layer is connected to the fully connected layer, and the fully connected layer is connected to the output layer, where N is a positive integer greater than or equal to 1; a rectified linear activation function is set after each convolutional layer; a normalized exponential function is set after the output layer. When constructing the network structure of the feature learning model, it is also necessary to set the initial parameters of the model, including parameters describing the network structure such as the number of convolutional / fully connected / downsampling layers, the number of convolutional kernels, and the size of the convolutional kernels. Convolutional neural networks have good fault tolerance, self-learning ability, and parallel processing ability for feature extraction and classification, allowing sampled data to have large defects and distortions, with fast running speed and good adaptive performance, which can reduce the complexity of the model, and its downsampling further reduces the output number of parameters and enhances the generalization ability of the model.

[0051] Exemplarily, Figure 2 is a schematic diagram of the network structure of a feature learning model, as Figure 2 shown. For the input data in the sample space, first, it is processed by 1 convolutional layer, where the convolutional kernel size is 3*3, the stride is 1, and the depth of the convolutional kernel is 32, for further mining and extraction of local features. Then, it passes through 1 2*2 pooling layer, using the max pooling method for dimensionality reduction processing of the feature output of the convolutional layer. After that, it goes through another convolutional layer and pooling layer to further extract features, reduce the model size, and increase the feature robustness. The activation function after each convolution operation selects the rectified linear function (Rectified Linear Unit, ReLu), and the function form is as follows:

[0052] f(x) = max(0, x)

[0053] where x is the input value and f(x) is the output value of the ReLU function. It can be seen that when the input value is greater than or equal to 0, the output of the ReLU function is the input value itself; when the input value is less than 0, the output of the ReLU function is 0. Using the ReLu activation function can make the expression ability of the neural network stronger, prevent the gradient disappearance problem, and accelerate the training speed.

[0054] The output of the last pooling layer passes through the fully connected layer to make the multi-dimensional input one-dimensional and integrate the features extracted by the CNN. Finally, the normalized exponential function (Softmax function) is selected as the output layer to output the cause analysis result, and the function form is as follows:

[0055]

[0056] Among them, z is a vector containing K elements, and z i is the data of the i-th element, and S i is the probability value of the i-th element. What is finally obtained by using the Softmax function is the probability distribution of each risk factor category. Therefore, the risk factor category with the highest probability is the prediction result output by the model.

[0057] S104: Use the sample data as input to train the feature learning model to obtain a risk assessment model.

[0058] Specifically, the stochastic gradient descent (SGD) algorithm can be used for network training. Exemplarily, the model training process includes the following steps: Input the feature data in the sample data into the feature learning model to obtain the predicted risk factor category, and calculate the error between the actual risk factor category and the predicted risk factor category as the loss value. Update the model parameters of the feature learning model according to this loss value and continue training until the loss value is less than the preset loss threshold. Among them, the model parameters are learnable parameters, which mainly include the weights and biases of each component.

[0059] Generally speaking, when training the model, the sample data will be divided into a training set and a test set according to a preset ratio. The training set is used to train the feature learning model. After the training is completed, the data in the test set is input into the trained model. The data in the test set is new data for the model. Therefore, it can detect the performance of the model on new data and comprehensively evaluate the performance of the model in various situations, including its generalization ability and stability. When the generalization ability and stability of the model are good, it can be finally determined that the model training meets the requirements. Otherwise, the model needs to be retrained.

[0060] S105: Use the real-time feature data of the road as input, and use the risk assessment model to predict the risk factor category of the road.

[0061] After obtaining the risk assessment model, the risk assessment model can be used to conduct a risk assessment on the real-time road, that is, the feature data of the road is collected in real time and input into the risk model to output the risk factor category of the road.

[0062] It should be noted that the feature data of roads in different regions and types may be different. If a certain item of feature data is not included in the feature data obtained by the current road, the value of this item of feature data can be a null value or 0, which can be specifically set according to actual needs.

[0063] A traffic risk factor assessment method provided by an embodiment of the present invention obtains initial data from historical traffic data. The initial data includes characteristic data of accident roads and actual risk factor categories that cause traffic risks; abnormal data screening and normalization processing are performed on the initial data to obtain sample data; a feature learning model is constructed using a deep learning algorithm; the feature learning model is trained with the sample data as input to obtain a risk assessment model; a feature learning model is constructed using a deep learning algorithm, and a machine learning model capable of identifying different risk characteristics is trained in combination with historical sample data, so as to quickly and accurately evaluate and analyze the traffic risk factor categories of different intersection sections, and can specifically guide the development of accident risk prevention work on actual roads, and timely prevent and control traffic risks.

[0064] Embodiment 2

[0065] Figure 3 It is a flowchart of a traffic risk factor assessment method provided by Embodiment 2 of the present invention. Embodiment 2 of the present invention is optimized on the basis of the above Embodiment 1. As Figure 3 shown, the traffic risk factor assessment method includes:

[0066] S301. Obtain initial data from historical traffic data. The initial data includes characteristic data of accident roads and actual risk factor categories that cause traffic risks.

[0067] S302. Perform abnormal data screening and normalization processing on the initial data to obtain sample data.

[0068] S303. Construct a feature learning model using a deep learning algorithm.

[0069] S301 - S303 are similar to S101 - S103 in Embodiment 1, and the specific description can refer to the relevant description of Embodiment 1.

[0070] S304. Input the characteristic data in the sample data into the feature learning model to obtain the predicted risk factor category.

[0071] S305. Calculate the loss function value according to the actual risk factor category and the predicted risk factor category.

[0072] Specifically, first obtain the probability distributions of the actual risk factor category and the predicted risk factor category according to the actual risk factor category and the predicted risk factor category respectively, and then calculate the loss function value. The calculation formula of the loss function value is:

[0073]

[0074] Among them, L m is the loss function value, N is the total number of the sample data, S is the total number of risk factor categories, ykc is the probability distribution of the actual risk factor category, p kc is the probability distribution of the predicted risk factor category.

[0075] S306. Determine whether the loss function value is less than a preset loss threshold.

[0076] If not, execute S307; if so, execute S308. Herein, the loss threshold can be set to 0.001.

[0077] S307. Update the parameters of the feature learning model based on the loss function value.

[0078] After executing S307, return to S304.

[0079] S308. Evaluate the model effect of the trained feature learning model.

[0080] Specifically, the model effect can be evaluated by calculating the Receiver Operating Characteristic (ROC) curve and the Area Under Curve (AUC) value, the confusion matrix, the F1 score, etc.

[0081] Optionally, evaluating the model effect of the trained feature learning model includes:

[0082] Using the trained feature learning model to draw the Receiver Operating Characteristic curve; calculating the Receiver Operating Characteristic curve; when the Area Under Curve value is greater than or equal to a preset area threshold, determining that the feature learning model fails the evaluation; when the Area Under Curve value is less than the preset area threshold, determining that the feature learning model passes the evaluation.

[0083] S309. Determine whether the feature learning model passes the evaluation.

[0084] If not, execute S310; if so, execute S311.

[0085] S310. Adjust the learning rate and dropout rate of the feature learning model.

[0086] After executing S310, return to S304.

[0087] The learning rate of the model has an important impact on the model training speed and training effect, while the dropout rate plays an important role in preventing model overfitting. In this embodiment, the initial learning rate is 0.001 and the initial dropout rate is 0.5. Exemplarily, for the trained evaluation model, AUC is used to evaluate the model effect. If the AUC value is greater than or equal to 0.8, it indicates that the model learning effect is good. If the AUC value is less than 0.8, the model learning rate and dropout rate parameters are adjusted, and the model is continuously trained until the optimal model is obtained.

[0088] In an alternative embodiment, the feature learning model includes a convolutional layer and a pooling layer. When the feature learning model fails to pass the evaluation, it further includes: determining whether the number of training times of the feature learning model reaches a preset number threshold; if so, increasing the number of convolutional layers and pooling layers in the feature learning model.

[0089] When the number of training times of the feature learning model reaches the preset number threshold, it means that the number of training times is large but the accuracy of the model is still difficult to meet the standard. Therefore, the number of convolutional layers and pooling layers in the feature learning model can be appropriately increased to improve the model accuracy. Since increasing the number of convolutional layers and pooling layers will complicate the model network structure, increase the training duration and computing power resources, and may also cause overfitting. Therefore, in the initial stage of model training (when the number of training times has not reached the preset number threshold), the increase in the number of convolutional layers and pooling layers is not considered. Only when the number of training times is large and the model accuracy is difficult to meet the standard, the number of convolutional layers and pooling layers is increased to meet the accuracy requirements of the model.

[0090] Specifically, when increasing the number of convolutional layers and pooling layers in the feature learning model, 1 convolutional layer and 1 pooling layer are added each time, and the settings of the newly added convolutional layer and pooling layer are the same as those of the existing convolutional layer and pooling layer, respectively.

[0091] S311. Use the current feature learning model as the risk assessment model.

[0092] S312. Use the real-time road feature data as the input, and use the risk assessment model to predict the risk factor category of the road.

[0093] In this embodiment, by setting the loss value, the calculation of AUC and the corresponding thresholds during the training of the risk assessment model, the model accuracy is fully guaranteed, so that the trained risk assessment model can accurately identify the risk factor category of the road for different feature data, that is, it can analyze and identify the risk factor category for different regions and different types of urban road traffic environments, and specifically guide the development of accident risk prevention work on actual roads, which is conducive to timely prevention and control of risk hazards.

[0094] Embodiment III

[0095] Figure 4 This is a schematic structural diagram of a traffic risk factor assessment device provided in Embodiment 3 of the present invention. As Figure 4 shown, the traffic risk factor assessment device includes:

[0096] An initial data acquisition module 100, configured to acquire initial data from historical traffic data, where the initial data includes characteristic data of accident roads and actual risk factor categories causing traffic risks;

[0097] A data processing module 200, configured to perform abnormal data screening and normalization processing on the initial data to obtain sample data;

[0098] A model construction module 300, configured to construct a feature learning model using a deep learning algorithm;

[0099] A model training module 400, configured to use the sample data as input to train the feature learning model to obtain a risk assessment model;

[0100] A model application module 500, configured to use the characteristic data of real-time roads as input and adopt the risk assessment model to identify the traffic risk factor categories of roads.

[0101] Optionally, the data processing module 200 includes:

[0102] An outlier processing sub-module, configured to detect outliers in the initial data by using the method of drawing a box plot and delete the outliers;

[0103] A data conversion sub-module, configured to convert non-numerical data in the initial data into numerical data according to the data type of the non-numerical data;

[0104] A data classification sub-module, configured to classify the initial data when the initial data are all numerical data;

[0105] A sample data determination sub-module, configured to perform normalization processing on each category of the initial data to obtain sample data.

[0106] Optionally, the feature learning model includes N convolutional layers, N pooling layers, 1 fully connected layer, and 1 output layer. During the training process, the output end of each convolutional layer is connected to the input end of 1 pooling layer, the output end of the last pooling layer is connected to the fully connected layer, and the fully connected layer is connected to the output layer.

[0107] Wherein, N is a positive integer greater than or equal to 1; a rectified linear activation function is provided after each convolutional layer; a softmax function is provided after the output layer.

[0108] Optionally, the model training module 400 includes:

[0109] A data input sub-module for inputting the feature data in the sample data into the feature learning model to obtain a predicted risk factor category;

[0110] A loss function value calculation sub-module for calculating a loss function value according to the actual risk factor category and the predicted risk factor category;

[0111] A loss function value judgment sub-module for judging whether the loss function value is less than a preset loss threshold;

[0112] A first parameter update sub-module for, if the loss function value is greater than or equal to the preset loss threshold, updating the parameters of the feature learning model based on the loss function value and returning to execute the content of the data input sub-module;

[0113] An effect evaluation sub-module for, if the loss function value is less than the preset loss threshold, evaluating the model effect of the trained feature learning model;

[0114] A second parameter update sub-module for, when the feature learning model fails the evaluation, adjusting the learning rate and dropout rate of the feature learning model and returning to execute the content of the data input sub-module;

[0115] A model determination sub-module for, when the feature learning model passes the evaluation, using the current feature learning model as the risk assessment model.

[0116] Optionally, the loss function value judgment sub-module includes:

[0117] A probability distribution calculation unit for obtaining the probability distributions of the actual risk factor category and the predicted risk factor category respectively according to the actual risk factor category and the predicted risk factor category;

[0118] A loss function value calculation unit for calculating a loss function value according to the probability distributions of the actual risk factor category and the predicted risk factor category, and the calculation formula of the loss function value is:

[0119]

[0120] where L m is the loss function value, N is the total number of the sample data, S is the total number of risk factor categories, y kc is the probability distribution of the actual risk factor category, and p kc is the probability distribution of the predicted risk factor category.

[0121] Optionally, the effect evaluation sub-module includes:

[0122] A characteristic curve drawing unit for drawing a receiver operating characteristic curve by using the trained feature learning model;

[0123] An area under the curve value calculation unit for calculating the area under the curve value of the receiver operating characteristic curve;

[0124] A first evaluation unit for determining that the feature learning model fails the evaluation when the area under the curve value is greater than or equal to a preset area threshold;

[0125] A second evaluation unit for determining that the feature learning model fails the evaluation when the area under the curve value is less than the preset area threshold.

[0126] Optionally, the feature learning model includes a convolutional layer and a pooling layer, and the second parameter update sub-module further includes:

[0127] A training times judgment unit for judging whether the training times of the feature learning model reach a preset times threshold when the feature learning model fails the evaluation;

[0128] A network structure update unit for increasing the number of convolutional layers and pooling layers in the feature learning model if the training times of the feature learning model reach the preset times threshold.

[0129] The traffic risk factor assessment device provided by the embodiments of the present invention can execute the traffic risk factor assessment method provided by any embodiment of the present invention, and has the corresponding functional modules and beneficial effects for executing the method.

[0130] Embodiment 4

[0131] Figure 5 FIG. shows a schematic structural diagram of an electronic device 40 that can be used to implement the embodiments of the present invention. The electronic device is intended to represent various forms of digital computers, such as, laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as, personal digital processors, cellular phones, smart phones, wearable devices (such as helmets, glasses, watches, etc.) and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the present invention described and / or claimed herein.

[0132] As Figure 5As shown, the electronic device 40 includes at least one processor 41 and a memory communicatively connected to the at least one processor 41, such as a read-only memory (ROM) 42, a random access memory (RAM) 43, etc. The memory stores a computer program executable by the at least one processor. The processor 41 can perform various appropriate actions and processes according to the computer program stored in the read-only memory (ROM) 42 or the computer program loaded from the storage unit 48 into the random access memory (RAM) 43. In the RAM 43, various programs and data required for the operation of the electronic device 40 can also be stored. The processor 41, the ROM 42, and the RAM 43 are connected to each other via a bus 44. An input / output (I / O) interface 45 is also connected to the bus 44.

[0133] Multiple components in the electronic device 40 are connected to the I / O interface 45, including: an input unit 46, such as a keyboard, a mouse, etc.; an output unit 47, such as various types of displays, speakers, etc.; a storage unit 48, such as a magnetic disk, an optical disc, etc.; and a communication unit 49, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 49 allows the electronic device 40 to exchange information / data with other devices via a computer network such as the Internet and / or various telecommunication networks.

[0134] The processor 41 can be various general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of the processor 41 include but are not limited to a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, a digital signal processor (DSP), and any appropriate processor, controller, microcontroller, etc. The processor 41 executes the various methods and processes described above, such as the traffic risk factor assessment method.

[0135] In some embodiments, the traffic risk factor assessment method can be implemented as a computer program tangibly embodied in a computer-readable storage medium, such as the storage unit 48. In some embodiments, part or all of the computer program can be loaded and / or installed onto the electronic device 40 via the ROM 42 and / or the communication unit 49. When the computer program is loaded into the RAM 43 and executed by the processor 41, one or more steps of the traffic risk factor assessment method described above can be executed. Alternatively, in other embodiments, the processor 41 can be configured to execute the traffic risk factor assessment method by any other appropriate means (e.g., by means of firmware).

[0136] The various embodiments of the systems and techniques described above in this specification can be implemented in digital electronic circuitry, integrated circuit systems, field programmable gate arrays (FPGA), application specific integrated circuits (ASIC), application specific standard products (ASSP), systems on a chip (SOC), complex programmable logic devices (CPLD), computer hardware, firmware, software, and / or combinations thereof. These various embodiments can include: being implemented in one or more computer programs that are executable and / or interpretable on a programmable system including at least one programmable processor, which can be a special-purpose or general-purpose programmable processor that receives data and instructions from, and transmits data and instructions to, a storage system, at least one input device, and at least one output device.

[0137] The computer programs for implementing the methods of the present invention can be written in any combination of one or more programming languages. These computer programs can be provided to a processor of a general purpose computer, special purpose computer, or other programmable data processing apparatus, such that the computer programs, when executed by the processor, cause the functions / operations specified in the flowchart and / or block diagram to be implemented. The computer programs can be executed entirely on the machine, partly on the machine, as a stand-alone software package partly on the machine and partly on a remote machine or entirely on the remote machine or server.

[0138] In the context of the present invention, a computer-readable storage medium can be a tangible medium that can contain or store a computer program for use by or in connection with an instruction execution system, apparatus, or device. The computer-readable storage medium can include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. Alternatively, the computer-readable storage medium can be a machine-readable signal medium. More specific examples of the machine-readable storage medium would include an electrical connection based on one or more wires, a portable computer diskette, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or Flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0139] To provide interaction with a user, the systems and techniques described herein can be implemented on an electronic device having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and a pointing device (e.g., a mouse or a trackball) by which the user can provide input to the electronic device. Other kinds of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including acoustic input, voice input, or tactile input).

[0140] The systems and techniques described herein can be implemented in a computing system including backend components (e.g., as a data server), or a computing system including middleware components (e.g., an application server), or a computing system including frontend components (e.g., a user computer having a graphical user interface or a web browser through which the user can interact with an implementation of the systems and techniques described herein), or a computing system including any combination of such backend components, middleware components, or frontend components. The components of the system can be interconnected to each other by digital data communication in any form or medium (e.g., a communication network). Examples of communication networks include: local area network (LAN), wide area network (WAN), blockchain network, and the Internet.

[0141] The computing system can include a client and a server. The client and the server are generally far from each other and usually interact through a communication network. The client-server relationship is created by computer programs running on the respective computers and having a client-server relationship with each other. The server can be a cloud server, also known as a cloud computing server or a cloud host, which is a host product in the cloud computing service system and solves the defects of difficult management and weak business scalability existing in traditional physical hosts and VPS services.

[0142] It should be understood that the various forms of processes shown above can be used, with steps reordered, added, or deleted. For example, the steps recited in the present invention can be executed in parallel, sequentially, or in a different order, as long as the desired results of the technical solution of the present invention can be achieved, and no limitation is made herein.

[0143] The above specific embodiments do not constitute a limitation on the protection scope of the present invention. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A method for evaluating traffic risk factors, characterized in that: include: Acquiring initial data from historical traffic data, wherein the initial data includes characteristic data of the accident road and categories of actual risk factors causing traffic risks; Performing abnormal data screening and normalization processing on the initial data to obtain sample data; Use deep learning algorithms to build feature learning models; Taking the sample data as input, training the feature learning model to obtain a risk assessment model; The risk assessment model is used to identify the traffic risk factor categories of the road using real-time road feature data as input.

2. The method according to claim 1, characterized in that The initial data is subjected to abnormal data screening and normalization processing to obtain sample data, including: Perform outlier detection on the initial data by drawing a box plot, and delete the outliers; For the non-numeric data in the initial data, convert the non-numeric data into numeric data according to the data type of the non-numeric data; When the initial data are all numerical data, classifying the initial data; Each type of the initial data is normalized to obtain sample data.

3. The method according to claim 1, characterized in that The feature learning model includes N convolutional layers, N pooling layers, 1 fully connected layer and 1 output layer. During the training process, the output end of each convolutional layer is connected to the input end of the pooling layer, the output end of the last pooling layer is connected to the fully connected layer, and the fully connected layer is connected to the output layer. Wherein, N is a positive integer greater than or equal to 1; a linear rectification activation function is set after each convolution layer; and a normalized exponential function is set after the output layer.

4. The method according to claim 1, characterized in that The sample data is used as input to train the feature learning model to obtain a risk assessment model, including: Inputting the feature data in the sample data into the feature learning model to obtain a predicted risk factor category; Calculating a loss function value according to the actual risk factor category and the predicted risk factor category; Determine whether the loss function value is less than a preset loss threshold; If not, updating the parameters of the feature learning model based on the loss function value, and returning to the step of inputting the feature data in the training set into the feature learning model to obtain a predicted risk factor category; If so, evaluating the model effect of the trained feature learning model; When the feature learning model fails the evaluation, the learning rate and the discard rate of the feature learning model are adjusted, and the step of returning to input the feature data in the sample data into the feature learning model to obtain a predicted risk factor category; When the feature learning model passes the evaluation, the current feature learning model is used as a risk assessment model.

5. The method according to claim 4, characterized in that The calculating the loss function value according to the actual risk factor category and the predicted risk factor category includes: Obtaining probability distributions of the actual risk factor category and the predicted risk factor category according to the actual risk factor category and the predicted risk factor category, respectively; The loss function value is calculated according to the probability distribution of the actual risk factor category and the predicted risk factor category. The calculation formula of the loss function value is: Among them, L m is the loss function value, N is the total number of sample data, S is the total number of risk factor categories, y kc is the probability distribution of the actual risk factor category, p kc is the probability distribution of the predicted risk factor categories.

6. The method according to claim 4, characterized in that The trained feature learning model is evaluated for model effect, including: Using the trained feature learning model to draw a receiver operating characteristic curve; Calculating the area under the curve of the receiver operating characteristic curve; When the area under the curve is greater than or equal to a preset area threshold, determining that the feature learning model fails the evaluation; When the area under the curve is less than a preset area threshold, it is determined that the feature learning model fails the evaluation.

7. The method according to claim 4, characterized in that The feature learning model includes a convolution layer and a pooling layer, and when the feature learning model fails the evaluation, further includes: Determine whether the training times of the feature learning model reaches a preset times threshold; If so, increase the number of convolutional layers and pooling layers in the feature learning model.

8. A traffic risk factor assessment device, characterized in that: include: An initial data acquisition module, used to acquire initial data from historical traffic data, wherein the initial data includes characteristic data of accident roads and categories of actual risk factors causing traffic risks; A data processing module is used to perform abnormal data screening and normalization processing on the initial data to obtain sample data; Model building module, used to build feature learning models using deep learning algorithms; A model training module is used to train a feature learning model using the sample data as input to obtain a risk assessment model; The model application module is used to use the real-time road feature data as input and adopt the risk assessment model to identify the traffic risk factor category of the road.

9. An electronic device, characterized in that: The electronic device comprises: at least one processor; and a memory communicatively connected to the at least one processor; wherein, The memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can execute the traffic risk factor assessment method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a processor to implement the traffic risk factor assessment method according to any one of claims 1 to 7 when executed.

Images