Driving behavior similarity analysis method and system based on collaborative filtering algorithm

Abstract

The invention provides a driving behavior similarity analysis method and system based on a collaborative filtering algorithm, and belongs to the technical field of vehicle safe driving. The driving behavior data of a driver is acquired; preprocessing the driving behavior data; calculating a weight correction factor and similarity of the preprocessed driving behavior data, and determining a node set and a neighbor set of a driver based on the obtained similarity; mapping the node set and the neighbor set of the driver into a node graph based on a graph neural network, and forming a correlation matrix by using nodes on the node graph; reflecting the driving behavior similarity of the driver by using the correlation matrix; and importing the correlation matrix into a system interface to realize visualization of the driving behavior similarity. According to the method, accurate analysis of the similarity of the driving behaviors of the driver can be realized, and the driver correlation model which is higher in universality and more stereoscopic is established.

Description

Technical Field

[0001] The present invention belongs to the technical field of vehicle safe driving, and in particular relates to a driving behavior similarity analysis method and system based on a collaborative filtering algorithm. Background Art

[0002] The statements in this section merely provide background information related to the present invention and do not necessarily constitute prior art.

[0003] With the new round of scientific and technological progress, the automotive industry has also ushered in industrial changes. Electrification, networking, intelligence, and sharing are collectively referred to as the "new four modernizations", which have become the guiding direction of industrial reform. In the field of intelligent transportation, drivers, vehicles, road environments, and safe driving are closely related. Among these three elements, drivers are human-centered, and there are individual differences in operating behaviors, which are important subjective factors that can affect driving safety. Studies have shown that about 87% of traffic accidents are caused by human factors, such as drunk driving, illegal driving, fatigue driving, distracted driving, etc. Therefore, the study of driving behavior has important theoretical and application value for ensuring traffic safety.

[0004] Researchers at home and abroad have conducted extensive research on driving behavior, including analyzing individual differences in driving behavior, evaluating driver behavior norms, and predicting the probability of abnormal driving behavior. The earliest research on establishing a standard model for drivers can be traced back to the 1930s, when Gibson et al. established a driver model based on the vehicle's driving path; for the task of driving style classification and identification, domestic scholars have successively proposed driving behavior research programs based on intelligent algorithms, including neural networks, bionics, fuzzy control, and fusion algorithms. Li Gang's team proposed an online recognition model based on neural networks and an offline recognition model based on HMM for the driver's preference characteristics for steering wheel control, and verified the effect of the model on steering preference retention on a car simulation platform. Cheng Bo et al. from Tsinghua University also established a hierarchical mathematical description method of driving state-behavior-mode, and proposed a driving behavior aggressiveness index based on Stevens' power law, realizing a quantitative evaluation of the aggressiveness of driving behavior.

[0005] However, the inventors found that the existing driving behavior research methods have some technical problems, such as:

[0006] (1) Traditional driving behavior research methods focus too much on the driving behavior of a single driver and ignore the correlation between driving behaviors. Taking the vehicle acceleration curve as an example, the industry's tuning method relies heavily on the engineering experience of experts. The formulation of its tuning rules usually lacks understanding and learning of the driver's personal preferences and only uses a single driving behavior evaluation technology. Therefore, it is difficult to achieve accurate analysis of the similarity of drivers' driving behaviors.

[0007] (2) Every driver has his or her own driving habits. Since the existing driving behavior research methods only consider the direct relationship between the driver and his or her own driving habits, the evaluation model constructed using the driving habits cannot be perfectly applied to the driving behavior analysis of other drivers. Therefore, the existing driving behavior research methods generally have the problem of weak generalization ability. Summary of the invention

[0008] In order to overcome the deficiencies of the above-mentioned prior art, the present invention provides a driving behavior similarity analysis method and system based on collaborative filtering algorithm, which can realize accurate analysis of the driving behavior similarity of drivers and establish a more universal and three-dimensional driver correlation model.

[0009] To achieve the above objectives, one or more embodiments of the present invention provide the following technical solutions:

[0010] A first aspect of the present invention provides a driving behavior similarity analysis method based on a collaborative filtering algorithm.

[0011] A driving behavior similarity analysis method based on collaborative filtering algorithm, comprising:

[0012] Obtaining the driver's driving behavior data;

[0013] Preprocessing the driving behavior data;

[0014] Calculating the weight correction factor and similarity of the preprocessed driving behavior data, and determining the node set and neighbor set of the driver based on the obtained similarity; mapping the node set and neighbor set of the driver into a node graph based on a graph neural network, and forming a correlation matrix with each node on the node graph; and using the correlation matrix to reflect the driving behavior similarity of the driver;

[0015] The correlation matrix is ​​imported into the system interface to realize the visualization of driving behavior similarity.

[0016] Furthermore, the driving behavior data includes vehicle travel data, travel abnormality data and driver behavior abnormality data.

[0017] Furthermore, the vehicle driving data includes driving speed, mileage, energy consumption, acceleration time and rapid acceleration time; the abnormal travel data includes lane departure alarm, forward collision alarm, vehicle distance alarm and pedestrian collision alarm; the abnormal driver behavior data includes fatigue driving alarm, phone call alarm, smoking alarm and distracted driving alarm.

[0018] Furthermore, the preprocessing includes data cleaning and normalization operations.

[0019] Further, calculating the similarity of the driving behavior data includes: calculating the cosine similarity of the driving behavior data, and multiplying the obtained cosine similarity by a weight correction factor to obtain the similarity of the driving behavior data.

[0020] Furthermore, the driver's node set and neighbor set are mapped into a node graph based on the graph neural network, including: taking maximizing the log-likelihood function as the optimization goal of the graph neural network; converting the optimization goal into a first loss value and a second loss value for solving the driving behavior, and performing weighted summation on the first loss value and the second loss value; and using the weighted sum value as the comparative learning loss function of the neighbor set.

[0021] Furthermore, the driver's node set and neighbor set are mapped into a node graph by taking the minimum contrastive learning loss function as the optimal solution.

[0022] A second aspect of the present invention provides a driving behavior similarity analysis system based on a collaborative filtering algorithm.

[0023] Driving behavior similarity analysis system based on collaborative filtering algorithm, including:

[0024] The data acquisition module is configured to: acquire the driving behavior data of the driver;

[0025] A preprocessing module is configured to: preprocess the driving behavior data;

[0026] The driving behavior similarity calculation module is configured to: calculate the weight correction factor and similarity of the preprocessed driving behavior data, determine the node set and neighbor set of the driver based on the obtained similarity; map the node set and neighbor set of the driver into a node graph based on a graph neural network, and use the nodes on the node graph to form a correlation matrix; and use the correlation matrix to reflect the driving behavior similarity of the driver;

[0027] The driving behavior similarity display module is configured to: import the correlation matrix into the system interface to realize the visualization of driving behavior similarity.

[0028] A third aspect of the present invention provides a computer-readable storage medium having a program stored thereon, which, when executed by a processor, implements the steps in the driving behavior similarity analysis method based on a collaborative filtering algorithm as described in the first aspect of the present invention.

[0029] The fourth aspect of the present invention provides an electronic device, including a memory, a processor, and a program stored in the memory and executable on the processor, wherein when the processor executes the program, the steps in the driving behavior similarity analysis method based on the collaborative filtering algorithm as described in the first aspect of the present invention are implemented.

[0030] One or more of the above technical solutions have the following beneficial effects:

[0031] (1) The present invention first calculates the weight correction factor and similarity of the driving behavior data, and determines the node set and neighbor set of the driver based on the obtained similarity; then maps the node set and neighbor set of the driver into a node graph based on the graph neural network, and uses the nodes on the node graph to form a correlation matrix; finally, the correlation matrix is ​​used to reflect the similarity of the driver's driving behavior. The entire analysis process not only considers the driving behavior of a single driver, but also considers the correlation between the driving behaviors of the drivers. Therefore, the present invention can achieve accurate analysis of the similarity of the drivers' driving behaviors.

[0032] (2) The present invention approaches from the perspective of similarity analysis and obtains the driver node feature similarity matrix through cosine similarity calculation to help the model better capture the potential connection between drivers and driving behaviors; in addition, by introducing a graph neural network, the similarity matrix is ​​converted into a graph for analysis, making more effective use of the graph structure. Therefore, the present invention establishes a more universal and three-dimensional driver correlation model.

[0033] Advantages of additional aspects of the present invention will be given in part in the following description, and in part will become obvious from the following description, or will be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] The accompanying drawings in the specification, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

[0035] Figure 1 The present invention is a flowchart of a method for analyzing driving behavior similarity based on a collaborative filtering algorithm in a first embodiment of the present invention. DETAILED DESCRIPTION

[0036] It should be noted that the following detailed descriptions are exemplary and are intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those commonly understood by those skilled in the art to which the present invention belongs.

[0037] It should be noted that the terms used herein are for describing specific embodiments only and are not intended to be limiting of exemplary embodiments according to the present invention.

[0038] In the absence of conflict, the embodiments of the present invention and the features of the embodiments may be combined with each other.

[0039] Embodiment 1

[0040] This embodiment discloses a driving behavior similarity analysis method based on a collaborative filtering algorithm.

[0041] like Figure 1 As shown, a driving behavior similarity analysis method based on collaborative filtering algorithm includes:

[0042] Step S1, obtaining the driver's driving behavior data;

[0043] Step S2, preprocessing the driving behavior data;

[0044] Step S3, calculating the weight correction factor and similarity of the preprocessed driving behavior data, and determining the node set and neighbor set of the driver based on the obtained similarity; mapping the node set and neighbor set of the driver into a node graph based on a graph neural network, and using each node on the node graph to form a correlation matrix; using the correlation matrix to reflect the driving behavior similarity of the driver;

[0045] Step S4: import the correlation matrix into the system interface to realize the visualization of driving behavior similarity.

[0046] Based on the above process, the present invention can achieve accurate analysis of the similarity of the driving behaviors of drivers and establish a more universal and three-dimensional driver correlation model. To facilitate the understanding of the technical solution of the present invention, the specific implementation steps of the technical solution of the present invention are further explained and illustrated below.

[0047] Step S1: Acquire the driver's driving behavior data.

[0048] The driver's daily driving behavior data is the basis of driving behavior analysis, which can help identify abnormal driving behavior, select reasonable abnormal driving behavior evaluation indicators, and establish an abnormal driving behavior data set.

[0049] Furthermore, the present embodiment obtains the driver's driving behavior data through the ADAS device, the DSM device and the CAN bus; wherein the driving behavior data includes vehicle travel data, abnormal travel data and abnormal driver behavior data. Specifically, the present embodiment obtains the abnormal travel data in the driver's driving behavior data through the ADAS device, obtains the abnormal driver behavior data in the driver's driving behavior data through the DSM device, and obtains the vehicle travel data in the driver's driving behavior data through the CAN bus.

[0050] Furthermore, in the specific process of obtaining driving behavior data, ADAS equipment and DSM equipment directly upload when an alarm is detected; while the CAN bus is reflected in the vehicle uploaded data when it is determined that the vehicle has abnormal behavior. The driver's driving behavior data obtained in this embodiment is shown in Table 1:

[0051] Table 1 Driving behavior data table

[0052]

[0053]

[0054] As shown in Table 1, after obtaining the driver's driving behavior data, the vehicle driving data, the travel abnormality data and the driver behavior abnormality data are used as three parallel first-level indicators, and the second-level indicators are established respectively under these three first-level indicators. Specifically: the second-level indicators under the vehicle driving data include: driving speed, mileage, form energy consumption, acceleration time and rapid acceleration time; preferably, in this embodiment, the designed acceleration time is the time when the acceleration exceeds 5. The second-level indicators under the travel abnormality data include: lane departure alarm, forward collision alarm, vehicle distance too close alarm and pedestrian collision alarm. The second-level indicators under the driver behavior abnormality data include fatigue driving alarm, phone call alarm, smoking alarm and distracted driving alarm. In summary, the three first-level indicators cover a total of thirteen second-level indicators, which can provide a theoretical basis for determining the driver's driving tendency and the establishment of a comprehensive scoring model for driving behavior.

[0055] Step S2: pre-process the driving behavior data.

[0056] Since the abnormal behavior data collection process is based on the actual driving scene of the vehicle, the influence of interference factors such as poor signal, sensor failure, and body vibration will cause some data to be missing or obviously unreasonable. In order to improve the efficiency and quality of data mining and maximize the use of effective information in the data, it is necessary to preprocess the data. Considering that the acquired abnormal behavior data will contain some missing values, and the existence of null values ​​will have a certain impact on the computer analysis of the data, the preprocessing process of the present invention includes data cleaning and normalization operations:

[0057] Step S2-1: Clean invalid data.

[0058] Invalid data refers to data that affects the algorithm task due to data accuracy and authenticity caused by sensor anomalies or other factors. This type of data is usually characterized by unreasonable data values. For example, in the displayed behavior data, values ​​less than 0 should not appear; therefore, this embodiment sets the lower limit of all behavior data to 0. In addition, in actual data, the data uploaded by vehicles every day should be positive and continuous. Therefore, the continuity of the data for each vehicle going online and offline needs to be judged to eliminate abnormal data.

[0059] Step S2-2: data normalization.

[0060] Normalization is performed on the cleaned data. Specifically, due to the different scales of sensor data, the data needs to be normalized before training the model so that the data of each dimension can be distributed in a similar range. At the same time, since the maximum and minimum values ​​of the data are relatively unstable, it is more likely to cause the normalization result to be unstable. Therefore, this embodiment adopts Z-Score normalization, so that the normalized data can still conform to the standard normal distribution, which not only does not cause the loss of key information, but also improves the performance of the model. The Z-Score normalization model is expressed as:

[0061]

[0062] in, represents the standardized result, represents sample data, and Represent the mean and standard deviation of the n-th dimension data respectively.

[0063] Step S3, calculating the weight correction factor and similarity of the preprocessed driving behavior data, and determining the driver's node set and neighbor set based on the obtained similarity; mapping the driver's node set and neighbor set into a node graph based on a graph neural network, and using the nodes on the node graph to form a correlation matrix; using the correlation matrix to reflect the driver's driving behavior similarity.

[0064] In step S3, this embodiment makes improvements based on the collaborative filtering algorithm, and calculates the similarity by analyzing the effective driving behavior of the driver. At the same time, driving behavior will be continuously updated and increased over time, and the impact of time changes on the calculation process cannot be ignored. Based on this, the concept of time series is introduced; the behavior weight correction factor is integrated with the traditional similarity calculation formula. By combining the two concepts, an improved similarity algorithm based on time series is generated, which avoids the problem of inaccurate calculation results caused by neglecting time factors and weight factors. Finally, the interactive data is constructed into a graph by combining the graph neural network, and the effective representation of the nodes is learned by GNN to establish a node feature similarity matrix that can reveal the potential connection between the driver and the driving behavior, that is, the correlation matrix. Specifically, step S3 can be implemented by the following steps:

[0065] Step S3-1: Calculate the weight correction factor and similarity of the preprocessed driving behavior data, and determine the driver's node set and neighbor set based on the obtained similarity.

[0066] In the most basic cosine similarity calculation process, the weight of each driving behavior is kept consistent by default, but in actual driving behavior analysis, the importance of features is different. In this regard, based on the above problem, the present invention proposes a behavior weight correction factor pun(i), namely:

[0067]

[0068] Where pun(i) represents the behavior weight correction factor of driving behavior i, count i It indicates the number of drivers who have had behaviors related to this feature, and all indicates the number of drivers.

[0069] Taking into full consideration the various factors that may affect the similarity calculation (mainly the behavior weight), this embodiment integrates the above-mentioned improved schemes, introduces the behavior weight correction factor into the modified cosine similarity calculation method, and characterizes the difference in the similarity calculation results through the dynamic change of the weight. Specifically, calculating the similarity of driving behavior data includes: calculating the cosine similarity of driving behavior data, and multiplying the obtained cosine similarity by the weight correction factor to obtain the similarity of driving behavior data. Therefore, the improved similarity calculation formula can be expressed as:

[0070]

[0071] Among them, sim pun (i, j) represents the similarity calculation between driving behaviors i and j, sim(i, j) represents the cosine similarity calculation between driving behaviors i and j, and pun(c) represents the behavior weight correction factor of driving behavior c.

[0072] Driving behavior grows with time, so the time factor can be visualized and arranged to generate a set; a time series refers to a set of variables sorted by time. Each variable represents a value at a certain moment, and the set generated by integrating these moment values ​​is called a time series.

[0073] And, according to the above similarity calculation, find the driver's neighbor set N∈U, specifically: add neighbors with similarity higher than 0.82 to the driver's neighbor set; where N represents the neighbor set and U represents the driver set (i.e., the driver's node set). It is worth noting that repeated experiments have shown that dividing the number of neighbors with similarity higher than 0.82 into neighbor sets is the most appropriate and most closely connected.

[0074] Step S3-2, based on the graph neural network, the driver's node set and neighbor set are mapped into a node graph, and the nodes on the node graph are used to form a correlation matrix. The graph neural network uses a graph structure to represent the relationship between the driver and the driving behavior, and generates feature information of the target node through a message passing mechanism, so that each node can aggregate the information of its high-order neighbors. By combining the graph neural network to construct the interaction data into a graph, and using GNN to learn the effective representation of the nodes, a node feature similarity matrix can be established to reveal the potential connection between the driver and the driving behavior. Among them, the driver's neighbor set refers to the set of driver nodes that are not directly reachable on the node graph but have similar characteristics or preferences. Specifically, step S3-2 can be implemented by the following method:

[0075] Step S3-2-1: Based on the similarity calculation in S3-1, a driver similarity matrix M is constructed. The Sigmoid function is applied to the similarity matrix M for nonlinear transformation to obtain the transformation matrix H, namely:

[0076]

[0077] Among them, H m and M m Represent the mth row of matrix H and matrix M respectively.

[0078] For the transformation matrix H, H m It is converted into a probability distribution through the Softmax function, and the probability of node self-connection is set to zero to prevent itself from affecting the calculation, that is:

[0079]

[0080] Among them, P m represents the influence matrix of the driver node, H m Represents the mth row of the transformation matrix. After constructing the influence matrix P of the driver node, it is possible to extract each driver node c from the influence matrix P i Comprehensive impact score i :

[0081]

[0082] By analyzing the driver node c i Comprehensive impact score i , the corresponding key nodes can be clearly identified. According to the above method, from the key node c j Start and get the grouped driver node c i In contrastive learning, the key node c j As a positive sample, c i are considered as negative samples.

[0083] Step S3-2-2, construct the interaction data of the driver and the driving behavior into an interaction graph G = (V, E); wherein V = U∪I is the set of nodes, that is, the set of drivers and neighbor sets; E = {(u,i)|u∈U,i∈I} is the set of edges, that is, the set of driving behaviors.

[0084] The optimization objective of the graph neural network is to maximize the log-likelihood function; the log-likelihood function is specifically expressed as:

[0085]

[0086] Where p represents the probability density function, u represents the driver node, and e u represents the set of driver nodes, Θ represents the set of model parameters, c i represents the key node of the driver, G represents the interaction matrix, and the interaction matrix is ​​the interaction data between the driver and the driving behavior.

[0087] Step S3-2-3, converting the above optimization objective into a first loss value and a second loss value for solving the driving behavior; wherein the first loss function for calculating the first loss value is expressed as:

[0088]

[0089] in, represents the first loss value; u represents the driver node, c j That is, the positive sample obtained in step S3-2-1, c i That is, the negative sample obtained in step S3-2-1, τ represents the temperature hyperparameter of the softmax function. Similarly, the goal of driving behavior is the same, and the second loss function used to calculate the second loss value is expressed as:

[0090]

[0091] in, represents the second loss value; i represents the driving behavior node, c k That is, the positive sample obtained in step S3-2-1, c t That is, the negative sample obtained by S3-2-1, τ is the temperature hyperparameter of the softmax function.

[0092] The first loss value and the second loss value calculated by the first loss function and the second loss function are weighted and summed, and the weighted sum is used as the contrastive learning loss function of the driver's neighbor set. Specifically, the contrastive learning loss function can be expressed as:

[0093]

[0094] Among them, L p represents the contrastive learning loss function, and α represents a hyperparameter used for the weights of the two losses in contrastive learning.

[0095] Based on the above process, the contrastive learning loss function L p The minimum is taken as the optimal solution, and the best model is continuously iterated to obtain the driver correlation matrix.

[0096] Step S4: import the obtained correlation matrix into the system interface to realize the visualization of driving behavior similarity.

[0097] This embodiment is based on the SpringBoot rapid development framework when designing the front-end and back-end, which can quickly and independently create projects integrated with Spring and mainstream frameworks; at the same time, an embedded Servlet container is used, so there is no need to generate a WAR package; and there are a lot of automatic configurations in the framework, which greatly simplifies the development process. Data service is the basis for supporting the interaction between the front-end and back-end. In terms of data storage, the present invention mainly uses the MYSQL database and adopts the JSON data exchange format for front-end and back-end data exchange, thereby creating a system interface.

[0098] After the system interface is created, the driver correlation model (i.e. the obtained correlation matrix) is imported into the created system, and the driver correlation matrix for the entire fleet can be obtained, that is, the visualization of the similarity of driving behaviors can be achieved; according to the display of the driver correlation matrix, the administrator can intuitively understand the driving tendencies and driving safety levels of the drivers in his fleet.

[0099] Based on the above design, the present invention can achieve accurate analysis of the similarity of drivers' driving behaviors and establish a more universal and three-dimensional driver correlation model.

[0100] Embodiment 2

[0101] This embodiment discloses a driving behavior similarity analysis system based on a collaborative filtering algorithm.

[0102] Driving behavior similarity analysis system based on collaborative filtering algorithm, including:

[0103] The data acquisition module is configured to: acquire the driving behavior data of the driver;

[0104] A preprocessing module is configured to: preprocess the driving behavior data;

[0105] The driving behavior similarity calculation module is configured to: calculate the weight correction factor and similarity of the preprocessed driving behavior data, determine the node set and neighbor set of the driver based on the obtained similarity; map the node set and neighbor set of the driver into a node graph based on a graph neural network, and use the nodes on the node graph to form a correlation matrix; and use the correlation matrix to reflect the driving behavior similarity of the driver;

[0106] The driving behavior similarity display module is configured to: import the correlation matrix into the system interface to realize the visualization of driving behavior similarity.

[0107] Embodiment 3

[0108] The purpose of this embodiment is to provide a computer-readable storage medium.

[0109] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps in the driving behavior similarity analysis method based on a collaborative filtering algorithm as described in the first embodiment of the present disclosure.

[0110] Embodiment 4

[0111] The purpose of this embodiment is to provide an electronic device.

[0112] An electronic device includes a memory, a processor, and a program stored in the memory and executable on the processor. When the processor executes the program, the steps in the driving behavior similarity analysis method based on the collaborative filtering algorithm as described in the first embodiment of the present disclosure are implemented.

[0113] The steps involved in the apparatuses of the above embodiments 2, 3 and 4 correspond to the method embodiment 1, and the specific implementation methods can refer to the relevant description part of embodiment 1. The term "computer-readable storage medium" should be understood as a single medium or multiple media including one or more instruction sets; it should also be understood to include any medium that can store, encode or carry an instruction set for execution by a processor and enable the processor to execute any method in the present invention.

[0114] Those skilled in the art should understand that the modules or steps of the present invention described above can be implemented by a general-purpose computer device, or alternatively, they can be implemented by a program code executable by a computing device, so that they can be stored in a storage device and executed by the computing device, or they can be made into individual integrated circuit modules, or multiple modules or steps therein can be made into a single integrated circuit module for implementation. The present invention is not limited to any specific combination of hardware and software.

[0115] Although the above describes the specific implementation mode of the present invention in conjunction with the accompanying drawings, it is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art on the basis of the technical solution of the present invention without creative work are still within the scope of protection of the present invention.

Claims

1. A driving behavior similarity analysis method based on collaborative filtering algorithm, characterized in that: include: Obtaining the driver's driving behavior data; Preprocessing the driving behavior data; Calculating the weight correction factor and similarity of the preprocessed driving behavior data, and determining the node set and neighbor set of the driver based on the obtained similarity; Based on a graph neural network, the node set and neighbor set of the driver are mapped into a node graph, and each node on the node graph is used to form a correlation matrix; Using the correlation matrix to reflect the similarity of the drivers' driving behaviors; The correlation matrix is ​​imported into the system interface to realize the visualization of driving behavior similarity.

2. The driving behavior similarity analysis method based on collaborative filtering algorithm according to claim 1, characterized in that: The driving behavior data includes vehicle travel data, travel abnormality data and driver behavior abnormality data.

3. The driving behavior similarity analysis method based on collaborative filtering algorithm as claimed in claim 2, characterized in that: The vehicle driving data includes driving speed, mileage, energy consumption, acceleration time and rapid acceleration time; the abnormal travel data includes lane departure alarm, forward collision alarm, vehicle distance alarm and pedestrian collision alarm; the abnormal driver behavior data includes fatigue driving alarm, phone call alarm, smoking alarm and distracted driving alarm.

4. The driving behavior similarity analysis method based on collaborative filtering algorithm according to claim 1, characterized in that: The preprocessing includes data cleaning and normalization operations.

5. The driving behavior similarity analysis method based on collaborative filtering algorithm according to claim 1, characterized in that: Calculating the similarity of the driving behavior data includes: calculating the cosine similarity of the driving behavior data, and multiplying the obtained cosine similarity with a weight correction factor to obtain the similarity of the driving behavior data.

6. The driving behavior similarity analysis method based on collaborative filtering algorithm according to claim 1, characterized in that: The node set and neighbor set of the driver are mapped into a node graph based on a graph neural network, including: maximizing the log-likelihood function as the optimization target of the graph neural network; converting the optimization target into a first loss value and a second loss value for solving the driving behavior, and performing a weighted summation on the first loss value and the second loss value; and using the value after the weighted summation as the comparative learning loss function of the neighbor set.

7. The driving behavior similarity analysis method based on collaborative filtering algorithm according to claim 6, characterized in that: The optimal solution is the one with the minimum contrastive learning loss function, and the driver's node set and neighbor set are mapped into a node graph.

8. Driving behavior similarity analysis system based on collaborative filtering algorithm, characterized by: include: The data acquisition module is configured to: acquire the driving behavior data of the driver; A preprocessing module is configured to: preprocess the driving behavior data; A driving behavior similarity calculation module is configured to: calculate a weight correction factor and a similarity of the preprocessed driving behavior data, and determine a node set and a neighbor set of the driver based on the obtained similarity; Based on a graph neural network, the node set and neighbor set of the driver are mapped into a node graph, and each node on the node graph is used to form a correlation matrix; Using the correlation matrix to reflect the similarity of the drivers' driving behaviors; The driving behavior similarity display module is configured to: import the correlation matrix into the system interface to realize the visualization of driving behavior similarity.

9. A computer-readable storage medium having a program stored thereon, characterized in that: When the program is executed by a processor, the steps in the driving behavior similarity analysis method based on a collaborative filtering algorithm as described in any one of claims 1 to 7 are implemented.

10. An electronic device comprising a memory, a processor, and a program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the steps in the driving behavior similarity analysis method based on the collaborative filtering algorithm as described in any one of claims 1 to 7 are implemented.

Images