Traffic safety precision education system based on traffic participant user portraits

Through multi-dimensional data processing and chameleon clustering algorithm based on traffic participants' personalities, risks, and traffic, user portraits are constructed, and the accuracy of target selection and content selection in traffic safety education is solved, and the effectiveness of education is improved .

CN118628310BActive Publication Date: 2024-11-15ZHEJIANG POLICE COLLEGE
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411098315.4
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Priority Date
2023-08-21
Filing Date
2024-08-12
Publication Date
2024-11-15
Estimated Expiration
2044-08-12

AI Technical Summary

Technical Problem

Traffic safety education lacks accuracy in target selection, content selection and effect evaluation, resulting in unsatisfactory education results.

Method used

Based on the multi-dimensional data processing of traffic participants' three dimensions: "personality", "risk" and "traffic", the chameleon clustering algorithm is used to divide personality categories and risk levels, build user portraits, and provide accurate missionary solutions through the missionary plan matching module.

Benefits of technology

Accurate screening and education of key personnel has been achieved, and the efficiency and effectiveness of traffic safety education has been improved.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118628310B_ABST
    Figure CN118628310B_ABST
Patent Text Reader

Abstract

The present invention provides a precise traffic safety education system based on user portraits of traffic participants, which belongs to the field of traffic safety education and big data application. It includes: a multi-dimensional data processing module for traffic participants, which constructs indicators of three dimensions of "personality", "risk" and "traffic"; a personality category classification module, which divides personality categories and dangerousness levels according to the "personality" dimension indicators; a risk level classification module, which divides risk levels according to the "risk" dimension indicators; a potential risk level classification module, which divides potential risk levels according to the "traffic" dimension indicators; a traffic participant user portrait module, which divides traffic participants into multiple types of user portraits according to their personality categories, risk levels, and potential risk levels; and an education program matching module, which matches education programs for different types of user portraits. The present invention depicts user portraits from multiple dimensions, which has important practical significance for achieving precise education for different user portraits.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the fields of traffic safety education and big data application, and in particular to a precise traffic safety education system based on user portraits of traffic participants. Background Art

[0002] In the work of traffic safety education, the general public is often given a broad education, and the target selection is relatively blind. There is a lack of analysis and summary of key personnel of different traffic violations. Therefore, the concept of user portraits of traffic participants is used to accurately judge the objects of education to fill this gap. At the same time, in terms of education content, traditional education forms cannot accurately push to different personnel, and there is a lack of methods to match personnel characteristics and video characteristics. In addition, there is a lack of comprehensive and quantitative evaluation methods for the evaluation of the more abstract short video publicity effect, so it is difficult to accurately grasp the publicity effect and adjust the direction after the education. In general, the current traffic safety education is not accurate and systematic in target selection, content selection and effect evaluation, which ultimately leads to unsatisfactory education results. In the field of traffic safety education, there is a lack of a traffic safety precision education system that provides all-round, multi-angle, quantifiable and precise guidance in the content production, promotion and release of short videos. Summary of the invention

[0003] In order to solve the above problems, the present invention provides a precise traffic safety education system based on user portraits of traffic participants, which starts from the three dimensions of "personality", "risk" and "traffic" of traffic participants to realize a precise traffic safety education system with all-round, multi-angle and quantifiable precise guidance.

[0004] In order to achieve the above object, the technical solution adopted by the present invention is as follows:

[0005] A precise traffic safety education system based on traffic participant user portraits, including:

[0006] Traffic participant multi-dimensional data processing module, which is used to construct indicators in three dimensions: "personality", "risk" and "traffic" based on the basic personal information of traffic participants;

[0007] A personality category classification module is used to classify the "personality" dimension indicators of traffic participants into multiple personality categories using a chameleon clustering algorithm, and to classify the danger level according to the personality categories;

[0008] A risk level classification module is used to classify traffic participants into corresponding risk levels according to their "risk" dimension indicators;

[0009] A potential risk level classification module is used to classify traffic participants into corresponding potential risk levels according to their "traffic" dimension indicators;

[0010] Traffic participant user portrait module, which is used to classify traffic participants into multiple types of user portraits according to their personality categories, risk levels, and potential risk levels;

[0011] The education plan matching module is used to match education plans for different types of user portraits.

[0012] The beneficial effects of the present invention are as follows: the present invention starts from the three dimensions of "personality", "risk" and "traffic" of traffic participants, and characterizes user portraits based on the chameleon clustering algorithm, thereby achieving accurate screening of key personnel and constructing a complete set of accurate traffic safety education system, which has important practical significance for achieving accurate education for key personnel and improving the efficiency and effectiveness of traffic safety education. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] In order to more clearly present the specific implementation methods and steps of the present invention, the drawings required for the specific implementation methods and steps are briefly introduced below.

[0014] Figure 1 Shown is a schematic diagram of the overall implementation process of the present invention;

[0015] Figure 2 The figure shows a schematic diagram of information included in the basic personal information database of traffic participants of the present invention;

[0016] Figure 3 Shown is a schematic diagram of the chameleon clustering algorithm of the present invention;

[0017] Figure 4 The figure shows a schematic diagram of the process of merging subclusters in the chameleon clustering process of the present invention;

[0018] Figure 5 The figure shows a schematic diagram of the process of creating a user portrait of a traffic participant according to the present invention;

[0019] Figure 6 Shown is a schematic diagram of the dimensions of the traffic participant user portrait of the present invention;

[0020] Figure 7 The figure shows a schematic diagram of the key personnel identification process of the present invention;

[0021] Figure 8 The figure shows a schematic diagram of the population to which the teaching scheme of the present invention is applicable;

[0022] Fig. 9 Shown is a schematic diagram of matching key personnel with the precise education plan of the present invention. DETAILED DESCRIPTION

[0023] The present invention is described in detail below in conjunction with the specific embodiments of the present application and the corresponding drawings.

[0024] As attached Figure 1 As shown, the traffic safety precision education system based on traffic participant user portrait proposed by the present invention mainly includes:

[0025] The multi-dimensional data processing module of traffic participants is used to construct indicators of the three dimensions of "personality", "risk" and "traffic" according to the basic personal information of traffic participants, that is, to execute the following step 1;

[0026] A personality category classification module is used to classify the "personality" dimension indicators of traffic participants into multiple personality categories using a chameleon clustering algorithm, that is, to execute the following steps 2-3;

[0027] A risk level classification module is used to classify traffic participants into corresponding risk levels according to the "risk" dimension indicators of traffic participants, that is, to execute the following step 4;

[0028] A potential risk level classification module is used to classify traffic participants into corresponding potential risk levels according to the "traffic" dimension indicators of the traffic participants, that is, to execute the following step 5;

[0029] A traffic participant user portrait module is used to classify traffic participants into multiple types of user portraits according to their personality categories, risk levels, and potential risk levels, i.e., to execute the following steps 6-7;

[0030] The education program matching module is used to match education programs for different types of user portraits, that is, to execute the following steps 8-10.

[0031] The above modules work together to realize the precise traffic safety education system based on traffic participant user portraits as follows:

[0032] Step 1: Preprocessing the basic personal information data of traffic participants to obtain a database of basic personal information of traffic participants, wherein the database contains a number of numerical indicators based on the three dimensions of "personality", "risk" and "traffic" of basic personal information;

[0033] Step 2: Classify the numerical indicators of the "personality" dimension of traffic participants based on the Chameleon clustering algorithm;

[0034] Step 3: Number the clusters of the “personality” dimension after clustering;

[0035] Step 4: Classify the “risk” of traffic participants based on big data;

[0036] Step 5: Determine the potential risks of traffic participants in the "traffic" dimension based on the hierarchical analysis method;

[0037] Step 6: Construct user portraits of traffic participants based on the Chameleon clustering algorithm;

[0038] Step 7: Number the clusters of traffic participants after clustering;

[0039] Step 8: Screen key personnel based on traffic participant user portraits;

[0040] Step 9: Develop a traffic safety education program;

[0041] Step 10: Accurately match the education plan to key personnel based on the user portraits of traffic participants.

[0042] Step 1: Build a database of basic personal information of traffic participants

[0043] As attached Figure 2 As shown in the following table 1-1, data information of 10 initial indicators including gender, age, education level, occupation, number of traffic violations, number of traffic accidents, monitoring data, checkpoint data, mobile phone signaling, and payment information are collected from the three dimensions of "personality", "risk" and "traffic" of traffic participants, as shown in the following table 1-1.

[0044] Table 1-1 Basic personal information of traffic participants

[0045]

[0046] The initial indicators of the "Transportation" dimension are explained as follows:

[0047] "Monitoring data" refers to road traffic monitoring data, which refers to the characteristic information such as vehicle type, color, size, etc. obtained by analyzing various typical traffic monitoring video streams such as road TV monitoring, smart card gates, and electronic eyes at intersections.

[0048] "Toll station data" refers to vehicle driving data collected by toll station equipment installed on the road, mainly including vehicle entry and exit time, license plate number, vehicle model, color and other information, and personnel's ID number, name, photo, etc.

[0049] "Mobile phone signaling" refers to mobile phone signaling data, which refers to the data generated during the interaction between mobile phones and wireless communication networks. By collecting communication behavior data, including information such as time and location, it can restore the residence and movement behavior of traffic participants, and supplement the data in areas not covered by road traffic monitoring data and checkpoint data. It is comprehensive, continuous, multidimensional and complete.

[0050] "Payment information" refers to personal payment data information, which mainly includes two levels: personal information and payment transaction information. Personal information includes necessary data for payment services, such as identity verification information such as ID number, bank account, payment account, mobile phone number, communication address, bank card password, security code, and other identity verification information, as well as the user's facial features, fingerprints, voice and other biometric information; payment transaction information includes the user's transaction amount, transaction time, transaction location, transaction channel, transaction frequency, purchased goods and service categories, etc. Through payment information, it can reflect the use of public transportation (such as buses, subways, etc.) and shared travel (such as shared bicycles, etc.) by traffic participants. The acquisition of the above-mentioned "payment information" must be authorized by the user and limited to the scope required to complete the specific purpose. The data obtained should comply with relevant laws and regulations during the processing process, and reasonable security measures should be taken to ensure the confidentiality and security of personal data, follow public order and good customs, and avoid causing adverse effects or hindering public interests.

[0051] All the basic information values ​​(non-empty) of a traffic participant are called a complete data. Next, the collected data is processed as follows:

[0052] 1.1 Missing Value Handling

[0053] Hot deck imputation is used to fill in the missing values ​​in the personal information of each traffic participant. That is, for a traffic participant with missing information, a traffic participant with the highest degree of overlap is found in the complete data, and the missing values ​​are filled with the corresponding data of the traffic participant.

[0054] 1.2 Noise data processing based on box plot

[0055] Due to problems with data collection equipment, errors in the data entry process, errors in the data transmission process, etc., there may be noise data in the data set of basic personal information of traffic participants. The specific processing methods are as follows:

[0056] 1) Noise data identification

[0057] Manually screen outliers in the basic personal information values ​​of traffic participants.

[0058] 2) Noise data processing

[0059] For the outliers that are screened out, they are regarded as missing values ​​and processed according to the missing value processing method in 1.1.

[0060] 1.3 Preprocessing of initial indicators in the “traffic” dimension

[0061] The indicator of "monitoring data" can mainly reflect information such as the main travel time periods, travel modes, travel OD, and travel frequency of traffic participants.

[0062] The indicator of "checkpoint data" can mainly reflect the entry and exit time, vehicle model and other information of motor vehicles as well as driver information.

[0063] For the indicator of "mobile phone signaling", the mobile phone signaling data can accurately reflect the travel patterns of traffic participants such as travel time, travel transportation mode, travel OD, travel frequency, etc. after being processed by a series of models.

[0064] For the indicator of "payment information", it can reflect the situation of transportation participants using public transportation (such as buses, subways, etc.) and shared travel (such as shared bicycles, etc.) and other modes of transportation.

[0065] The present invention summarizes the travel characteristics reflected by the indicators of the "traffic" dimension of the above four traffic participants, and obtains four travel characteristics of the traffic participants, namely, the main travel time period, travel mode, travel OD, and travel frequency, and takes the travel characteristics as the final indicator of the "traffic" dimension.

[0066] Through the above processing, a data set of personal basic information of traffic participants containing 10 indicators is finally obtained, forming a database of personal basic information of traffic participants.

[0067] Step 2: Categorize traffic participants’ “personality” dimensions based on the Chameleon clustering algorithm

[0068] Since the dimension of "personality" includes four indicators, namely "gender", "age", "education" and "occupation", among which the "age" indicator is a numerical indicator, in order to facilitate quantitative statistical analysis, the remaining three indicators are numerically processed by assignment. The specific assignment method is as follows:

[0069] For the indicator “gender”, males were assigned a value of “1” and females were assigned a value of “2”.

[0070] For the indicator of "educational qualifications", the educational qualifications of primary school, junior high school, regular high school, technical school, vocational high school, secondary technical school, secondary normal school, higher technician, higher vocational college, college, undergraduate, master's degree and doctoral degree are assigned values ​​of "1" to "13" respectively.

[0071] For the indicator of “occupation”, according to the “Classification and Code of National Economic Industries”, the following 13 occupations are assigned values ​​of “1” to “13” in sequence: (1) agriculture, forestry, animal husbandry, fishery, and water conservancy; (2) industry; (3) geological survey and exploration; (4) construction; (5) transportation, postal and telecommunications; (6) commerce, public catering, material supply and warehousing; (7) real estate management, public utilities, resident services and consulting services; (8) health, sports and social welfare; (9) education, culture, art, and radio and television; (10) scientific research and comprehensive technical services; (11) finance and insurance; (12) government agencies and social groups; (13) other industries.

[0072] Place the dataset containing these four numerical indicators of all traffic participants in a 4-dimensional Euclidean space In the process, the Chameleon clustering algorithm is used for processing. The specific operation method is attached. Figure 3 shown.

[0073] 2.1 Select the nearest point

[0074] For the preprocessed traffic participant "personality" dimension information dataset, the dataset is denoted as ,in Indicates data objects, i.e., A complete set of data including "gender", "age", "education" and "occupation" for traffic participants. In The data objects correspond to the 4-dimensional Euclidean space In points.

[0075] The indicators "gender", "age", "education" and "occupation" are respectively indicators, all data objects Transformed into the following feature vector:

[0076]

[0077] in, Represents a data object No. A numerical indicator value.

[0078] Any two data objects The corresponding points are in 4-dimensional Euclidean space The Euclidean distance in for:

[0079]

[0080] For Data objects , respectively calculate this point and other The distances of the points are sorted from small to large, and the closest distance to the data object is selected. data objects, that is, the corresponding 4-dimensional Euclidean space In nearest neighbor points.

[0081] in, Selection and total number of data objects In order to prevent Too small or too large will cause too large an error, so this embodiment uses the empirical rule ( ) is used to determine the value.

[0082] from Initially, use the test set to estimate the error rate of the classifier, repeat the process, increase by 1 each time, allow 1 neighboring selected data object to be added, sort the error rates, and take the one that produces the smallest error rate. As the number of nearest neighboring data objects to be selected.

[0083] Recursive Dataset The points corresponding to all data objects in the data set until each point is determined separately. nearest neighbor points.

[0084] 2.2 Constructing KNN (K-nearest neighbor) graph

[0085] The present invention uses edge weight to represent the similarity between two data objects. The larger the Euclidean distance between the points corresponding to the two data objects, the smaller the edge weight and the smaller the similarity. Conversely, the smaller the Euclidean distance between the points corresponding to the two data objects, the larger the edge weight and the greater the similarity.

[0086] In this embodiment, each point obtained in 2.1 is compared with its The Euclidean distance of the nearest neighbor The inverse of ,Right now:

[0087]

[0088] As a point in space One of its nearest neighbors The edge weight is .

[0089] In 4-dimensional Euclidean space According to the data set The feature vector of each data object in the graph is plotted, and the node corresponding to the data object is connected to each node with a weighted edge. The nearest neighbor points are found, and the thickness of each edge is determined according to the size of the edge weight to form a KNN graph.

[0090] 2.3 Sub-clustering

[0091] According to the graph partitioning algorithm, the KNN graph is divided into a large number of unconnected smaller subgraphs, each of which represents an initial subcluster of hierarchical clustering.

[0092] The original KNN graph is called a cluster .

[0093] The specific operation process is as follows:

[0094] 1) Divide the KNN graph into two approximately equal subgraphs with a cut edge, so that the sum of the total edge weights of the cut edge is minimized. Divide into two subclusters The sum of the edge weights of the edges that need to be cut is the smallest. The approximately equal subgraphs refer to the two subgraphs having the same number of nodes or a difference of 1.

[0095] 2) Treat each subgraph as an initial cluster and repeat the above process until the number of nodes in each generated subgraph reaches a certain standard, and finally generate sub-clusters. The subclusters are placed in the list to be merged.

[0096] 2.4 Merging subclusters

[0097] As attached Figure 4 As shown, the present invention combines two clusters The relative interconnectivity between and relative similarity Measuring the similarity of subclusters obtained by different partitioning methods , according to the similarity With threshold The relationship between the sub-clusters determines whether to merge or not, thereby obtaining the final clustering result. The specific operation method is as follows:

[0098] 1) Determine the similarity function that measures the similarity between two clusters:

[0099] Define the relative interconnectedness between two clusters for:

[0100]

[0101] in, Representation Cluster Divided into two subclusters The weight of the cut edge, Respectively indicate that The weight of the cut edge that is divided into two approximately equal parts.

[0102] Define the relative proximity between two clusters for:

[0103]

[0104] in, Representation Cluster Divided into two subclusters The average weight of the cut edge, that is, the connection sub-cluster Nodes and subclusters in The average weight of the edges of the nodes in Respectively indicate that The average weight of the cut edge that is divided into two roughly equal parts, that is, the minimum bisection cluster (or )’s average weight; , Represent subclusters and sub-cluster The number of nodes in .

[0105] Introducing parameters To adjust the relative interconnectivity and relative similarity The ratio of these two parameters is Indicates that relative connectivity is more important. Indicates that relative similarity is more important. Therefore, the similarity function is defined as:

[0106]

[0107] Defining Thresholds (This threshold is assigned according to the specific situation of the collected data set).

[0108] 2) Process the subclusters in the list to be merged in order. For each subcluster, calculate its neighbor subclusters (where , Determined according to the number of sub-clusters divided in 2.3) and , the corresponding similarity is calculated through the similarity function ;

[0109] 3) For all the division methods indivual Sort from largest to smallest. If the largest Exceeding the threshold ,Right now , then the subcluster is associated with the largest The corresponding subclusters are merged;

[0110] 4) If the largest Not exceeding the threshold ,Right now , it means that the sub-cluster has been merged and the sub-cluster is removed from the list to be merged;

[0111] 5) Repeat steps 1) to 4) for the next sub-cluster in the list until the list to be merged is empty, that is, the clustering of all traffic participants in the data set is completed, and The final clusters ( determined according to a hierarchical clustering procedure).

[0112] Step 3: Number the clusters of the “Personality” dimension after clustering

[0113] For the obtained The final clusters are summarized according to the value of the most repeated indicator of the traffic participants in each cluster, such as "male retired (or unemployed) with low education level" and "female middle-aged civil servant with high education level". Traffic participants with more distinct characteristics are numbered 1~ Types of traffic participants can be reflected in Table 3-1 below.

[0114] Table 3-1 Category number table of traffic participants' "personality" dimension

[0115]

[0116] right The riskiness of a personality category can be judged based on age. For example, taking age as an example, a personality category containing age characteristics under 18 years old or age characteristics over 60 years old can be marked as high risk.

[0117] The definition of high risk includes but is not limited to the above situations. Those skilled in the art can also define the risk based on other specific features of the personality category or a combination of different specific features, and mark the personality category number as "high risk" or "low risk".

[0118] Step 4: Classify the “risk” of traffic participants based on big data

[0119] 4.1 Assignment of “risk” dimension to traffic participants

[0120] The value assigned to the dimension of "risk" in the present invention is determined by the accumulation of the two index values ​​of "violation" and "accident". "Violation" specifically refers to the traffic violations of traffic participants in a unit time, which is determined by the illegal behaviors of motor vehicle drivers, non-motor vehicle drivers, pedestrians, and passengers; "accident" specifically refers to the traffic accident situation of traffic participants in a unit time, which is jointly determined by the traffic accident responsibility level (minor accident, general accident, major accident, major accident) and traffic accident responsibility (full responsibility, main responsibility, equal responsibility, secondary responsibility, no responsibility). The specific operation method is as follows:

[0121] The present invention evaluates the risk of the traffic participant based on the traffic violations and accidents in a unit time. Let the unit time be ( It is determined according to the needs of different regions and periods, which can be six months, one year, two years, three years, etc.), and the violation records and accident data of each traffic participant in that unit time are retrieved from the relevant database.

[0122] 1) Traffic violation status assignment

[0123] For traffic violations committed by motor vehicle drivers, according to the scoring rules in the "Road Traffic Safety Violation Points Management Measures", the number of violations committed by the traffic participant that are worth 1 point, 3 points, 6 points, 9 points, and 12 points are recorded respectively. , then the traffic participant's motor vehicle violation points for:

[0124]

[0125] For non-motor vehicle drivers' traffic violations, the present invention assigns 2 points for each fine, and , determine the non-motor vehicle violation points for the traffic participant :

[0126]

[0127] For other traffic violations (including pedestrians, passengers, etc.), the present invention will record 1 point for each violation, and the number of violations will be calculated according to the number of violations. Determine other illegal points for the traffic participant :

[0128]

[0129] So the number is Assignment of traffic violations of traffic participants for:

[0130]

[0131] 2) Traffic accident situation assignment

[0132] For traffic accident situations involving traffic participants, points are assigned based on the identified traffic accident responsibility level (minor accident, general accident, major accident, major accident) and traffic accident responsibility (full responsibility, main responsibility, equal responsibility, secondary responsibility, no responsibility). The specific points are shown in Table 4-1 below:

[0133] Table 4-1 Traffic accident scoring table

[0134]

[0135] Combined with the above scoring table, according to the number of traffic accidents committed by traffic participants c Points are assigned according to the corresponding level and responsibility The product of is calculated as Assignment of traffic accident conditions to traffic participants for:

[0136]

[0137] Comprehensive traffic violations and traffic accidents, determine the number The assignment of the dimension of “risk” to traffic participants for:

[0138]

[0139] 4.2 “Risk” classification of traffic participants

[0140] The present invention divides the risk level of traffic participants into three levels: "low risk", "medium risk" and "high risk". For "low risk", For "medium risk", is “high risk”.

[0141] Step 5: Determine the potential risks of traffic participants in the "traffic" dimension based on the analytic hierarchy process

[0142] Since the four indicators of the "traffic" dimension are non-numerical indicators, in order to facilitate quantitative statistical analysis, they are numerically processed by assignment. The specific assignment method is as follows:

[0143] Table 5-1 Traffic travel characteristics assignment table

[0144]

[0145] The present invention defines "potential risk value" as: for traffic participants whose "risk" dimension is "medium risk" or "low risk", under certain travel time periods, traffic modes, travel OD, travel frequency and other conditions, the risk level of traffic violations or traffic accidents.

[0146] 5.1 Determining label weights based on analytic hierarchy process

[0147] Combined with the assignment of the travel characteristics of traffic participants in step 1 (see Table 5-1 for details), the importance of the four labels of traffic participants, namely, travel time, travel mode, travel OD, and travel frequency, is evaluated using the analytic hierarchy process (AHP). The specific process is as follows:

[0148] 5.1.1 Establishing a Hierarchical Structure Model

[0149] A hierarchical structure model is established with the decision-making objectives (travel characteristics of traffic participants), decision-making criteria (evaluation labels of travel characteristics, including travel time, travel transportation mode, travel OD, and travel frequency), and decision-making plans (15 specific situations in Table 5-1) as the highest layer, middle layer, and lowest layer, respectively.

[0150] 5.1.2 Setting the judgment scale

[0151] The travel time period, travel transportation mode, travel OD, and travel frequency are recorded as the 1st, 2nd, 3rd, and 4th evaluation labels respectively.

[0152] In order to facilitate the quantification of comparative judgment, a 1-9 ratio scaling method is introduced, and 1, 3, 5, 7, and 9 are used to represent the evaluation labels based on experience. With evaluation label The relative importance levels are: equally important, slightly important, relatively important, strongly important, and absolutely important, while 2, 4, 6, and 8 represent the intermediate values ​​between the above two judgment levels, resulting in a judgment scale as shown in Table 5-2.

[0153] Table 5-2 Judgment scale table

[0154]

[0155] Provision: To judge the matrix elements Indicates the evaluation label With evaluation label v The relative importance of Indicates the evaluation label v With evaluation label The relative importance of , then the two satisfy:

[0156]

[0157] Obviously, The larger the value, the more it indicates the evaluation label. The greater the importance; on the contrary, it means that the evaluation label The less important.

[0158] 5.1.3 Constructing the Judgment Matrix

[0159] Two experts in traffic safety management, especially traffic safety education, and two experts in the field of big data were invited to independently score the four evaluation labels of travel time, travel mode, travel OD, and travel frequency according to the judgment scale in Table 5-2. The evaluation label scoring table is shown in Table 5-3.

[0160] Table 5-3 Evaluation label scoring table

[0161]

[0162] According to the scores of the four experts, the average of the four scores is taken for each element of the judgment matrix to construct the importance judgment matrix of the evaluation labels of the travel characteristics of traffic participants:

[0163]

[0164] 5.1.4 Evaluation label weight calculation and consistency check

[0165] The indicator weight calculation and consistency test are performed on the above importance judgment matrix respectively. The specific process is as follows:

[0166] 1) The judgment matrix Normalize by column, let for After normalization, the variables are:

[0167]

[0168] 2) Sum the normalized matrix row by row, let is the row sum after summing, then:

[0169]

[0170] 3) Yes Perform normalization processing, let is the normalized variable, which is the evaluation label Weight:

[0171]

[0172] Get the feature vector :

[0173]

[0174] 4) Calculate the eigenvector The maximum eigenvalue of :

[0175]

[0176] in, Represents the first elements.

[0177] 5) Calculate consistency index :

[0178]

[0179] By querying the average random consistency index From the value table, we can see that when the order is 4, .

[0180] Perform consistency check on the judgment matrix and calculate the consistency ratio :

[0181]

[0182] If the consistency index , then the consistency requirement is met, indicating that the judgment matrix is ​​valid. At this time, the judgment matrix passes the preliminary test; if , then the consistency requirements are not met and the consistency test fails. At this time, the indicators of the judgment matrix are adjusted and the consistency test is performed again until it passes the consistency test.

[0183] 5.2 Calculating Potential Risk Value

[0184] According to the assignment of specific conditions under each evaluation label in Table 5-2, record the evaluation label The assignment value is .

[0185] The evaluation label weight after passing the consistency test The importance of the four evaluation labels as the travel characteristics of traffic participants is calculated by multiplying the value of each evaluation label by the importance of the evaluation label and adding them up to get the number Potential risk value of traffic participants:

[0186]

[0187] 5.3 Potential risk classification

[0188] The present invention divides the potential risk level of traffic participants into two levels: "low potential risk" and "high potential risk". The potential risk is low. The risk is “potentially high”.

[0189] Step 6: Constructing traffic participant user portraits based on the Chameleon clustering algorithm

[0190] User portraits generally include the user's social attributes, living habits, consumption behaviors and other main information. In the present invention, the personality category, risk level and potential risk level obtained according to the indicators of the three dimensions of "personality", "risk" and "traffic" of traffic participants are used to construct user portraits.

[0191] As attached Figure 5 As shown, the present invention uses "personality category", "risk level" and "potential risk level" information as characteristic information of traffic participants, and uses the chameleon clustering algorithm to cluster traffic participants, so the three-dimensional data set containing all traffic participants is placed in a 3-dimensional Euclidean space Processing, such as Figure 6 The specific operation method is as follows:

[0192] 6.1 Selecting the nearest neighbor

[0193] For the preprocessed traffic participants’ personal basic information dataset, the dataset is denoted as ,in Indicates data objects, i.e., A complete set of data on traffic participants, including information on three dimensions: "personality category", "risk level", and "potential risk level". In Each data object corresponds to a 3-dimensional Euclidean space In nodes.

[0194] Note that “Personality Category”, “Risk Level” and “Potential Risk Level” are the dimensions, all data objects Transformed into the following feature vector:

[0195]

[0196] in Represents a data object No. An indicator value.

[0197] Any two data objects The corresponding points are in 3D Euclidean space The Euclidean distance in for:

[0198]

[0199] For Data objects , respectively calculate this point and other The distances of the points are sorted from small to large, and the closest distance to the data object is selected. data objects, that is, the corresponding 3D Euclidean space In nearest neighbor points.

[0200] in, Selection and total number of data objects In this invention, the empirical rule ( ) is used to determine the value.

[0201] from Initially, use the test set to estimate the error rate of the classifier, repeat the process, increase by 1 each time, allow 1 neighboring selected data object to be added, sort the error rates, and take the one that produces the smallest error rate. As the number of nearest neighboring data objects to be selected.

[0202] Recursive Dataset The points corresponding to all data objects in the data set until each point is determined separately. nearest neighbor nearest neighbor.

[0203] 6.2 Constructing KNN (K-nearest neighbor) graph

[0204] In this embodiment, each point obtained in 6.1 is compared with its The Euclidean distance of the nearest neighbor The inverse of ,Right now:

[0205]

[0206] As a point in space One of its nearest neighbors The edge weight is .

[0207] In 3D Euclidean space According to the data set The feature vector of each data object in the graph is plotted, and the node corresponding to the data object is connected to each node with a weighted edge. The nearest neighbor points are found, and the thickness of each edge is determined according to the size of the edge weight to form a KNN graph.

[0208] 6.3 Sub-clustering

[0209] According to the graph partitioning algorithm, the KNN graph is divided into a large number of unconnected smaller subgraphs, each of which represents an initial subcluster of hierarchical clustering. The original KNN graph is called a cluster .

[0210] The specific operation process is as follows:

[0211] 1) Divide the KNN graph into two approximately equal subgraphs with a cut edge, so that the sum of the total edge weights of the cut edge is minimized. Divide into two subclusters The sum of the edge weights of the edges that need to be cut is the smallest. The approximately equal subgraphs refer to the two subgraphs having the same number of nodes or a difference of 1.

[0212] 2) Treat each subgraph as an initial cluster and repeat the above process until the number of nodes in each generated subgraph reaches a certain standard, and finally generate sub-clusters. The subclusters are placed in the list to be merged.

[0213] 6.4 Merging Subclusters

[0214] The present invention combines two clusters The relative interconnectivity between and relative similarity Measuring the similarity of subclusters obtained by different partitioning methods , according to the similarity With threshold The relationship between the sub-clusters determines whether to merge or not, thereby obtaining the final clustering result. The specific operation method is as follows:

[0215] 1) Determine the similarity function that measures the similarity between two clusters:

[0216] The relative interconnectedness between two clusters for:

[0217]

[0218] in, Representation Cluster Divided into two subclusters The weight of the cut edge, Respectively indicate The weight of the cut edge that is divided into two approximately equal parts.

[0219] The relative proximity between two clusters for:

[0220]

[0221] in, Representation Cluster Divided into two subclusters The average weight of the cut edge, that is, the connection sub-cluster Nodes and subclusters in The average weight of the edges of the nodes in Respectively indicate that The average weight of the cut edge that is divided into two roughly equal parts, that is, the minimum bisection cluster (or )’s average weight; , Represent subclusters and sub-cluster The number of nodes in .

[0222] Introducing parameters To adjust the relative interconnectivity and relative similarity The ratio of these two parameters is Indicates that relative connectivity is more important. Indicates that relative similarity is more important. Therefore, the similarity function for:

[0223]

[0224] Defining Thresholds (This threshold is assigned according to the specific situation of the collected data set).

[0225] 2) Process the subclusters in the list to be merged in order. For each subcluster, calculate its neighbor subclusters (where , Determined according to the number of sub-clusters divided in 6.3) and , the corresponding similarity is calculated through the similarity function ;

[0226] 3) For all the division methods indivual Sort from largest to smallest. If the largest Exceeding the threshold ,Right now , then the subcluster is associated with the largest The corresponding subclusters are merged;

[0227] 4) If the largest Not exceeding the threshold ,Right now , it means that the sub-cluster has been merged and the sub-cluster is removed from the list to be merged;

[0228] 5) Repeat steps 1) to 4) for the next sub-cluster in the list until the list to be merged is empty, that is, the three-dimensional clustering of all traffic participants in the data set is completed, and The final clusters ( determined according to a hierarchical clustering procedure).

[0229] Step 7: Number the clusters of traffic participants after clustering

[0230] For the obtained The final clusters are obtained by summarizing the information of the three dimensions of traffic participants in each cluster, such as "Personality No. 1 (male, old, low-educated, retired), high risk, high potential risk", "Personality No. 2 (female, middle-aged, highly educated, civil servant), low risk, low potential risk", etc. The traffic participants with distinct characteristics are obtained, that is, the user portraits of each type of traffic participants are obtained, and they are numbered from 1 to Types of traffic participants can be reflected in Table 7-1 below.

[0231] Table 7-1 Traffic Participant Category Number Table

[0232]

[0233] Step 8: Identification of key personnel for traffic safety education based on traffic participant user portraits

[0234] In order to fully and reasonably allocate existing resources for precise education, the first By analyzing the three dimensions and multiple indicators of user portraits, we can identify key personnel for traffic safety education with higher traffic safety risks, such as Figure 7 shown.

[0235] 8.1 Key personnel with “high risk” in the “risk” dimension

[0236] Among all the categories of traffic participants, according to the indicators of the dimension of "risk", the personnel are divided into "high risk", "medium risk" and "low risk". For those who are classified as "high risk", they have committed traffic violations many times, or have committed serious traffic violations and have been responsible for more than one serious accident, which means that they have more aggressive driving habits, lack of safety awareness, and low compliance with traffic laws and regulations. Such personnel have a high degree of traffic safety risk and are judged as key personnel with "high risk".

[0237] 8.2 Key Personnel with “High Potential Risk” in the “Transportation” Dimension

[0238] For those who are classified as "medium risk" or "low risk" in the "risk" dimension, they are classified from the "transportation" dimension. For those who have high risk in the main travel time, travel mode, travel OD, and high average travel frequency, they are identified as key personnel with "high potential risk". Since the main modes of transportation for this type of personnel are large trucks, large / medium-sized operating buses, special road vehicles for transporting dangerous chemicals, fireworks, and civilian explosives, etc., and due to work and other reasons, these people need to drive special vehicles that are prone to accidents or cause serious consequences in dangerous periods and dangerous paths for a long time. The average travel frequency is high, and in the same traffic scenario, the risk coefficient of accidents is higher, so the personnel with "medium risk" or "low risk" in the "risk" dimension and "high potential risk" in the "transportation" dimension are identified as key personnel.

[0239] 8.3 Key personnel with “high risk” in the “personality” dimension

[0240] For people who are classified as "medium risk" or "low risk" in the "risk" dimension, and whose main travel time, travel mode, and travel OD do not belong to the type with a high risk factor in the "transportation" dimension, but who have a high travel frequency, the present invention compares the "personality" characteristics of the key personnel identified in 8.1 and 8.2, finds people with similar characteristics, and identifies them as key personnel in the "personality" dimension. The characteristics of this part of people are mainly manifested in poor traffic safety awareness and low compliance with traffic laws and regulations.

[0241] Therefore, among all the categories of traffic participants, based on the indicators of the dimension of "personality" of the personnel, the traffic participants who have not been identified as key persons in 8.1 and 8.2 will continue to be identified. If their personality category corresponds to "high risk", they will be identified as key persons with "high risk".

[0242] The "personality" dimension, as a supplement to the "risk" and "traffic" dimensions, has a significant effect in detecting omissions. For example, for a minor traffic participant under the age of 18, this type of person is a minor, and the main mode of travel in the "traffic" dimension is non-motorized vehicles and the travel frequency is high. According to the standard of 8.2, it is highly likely that they will not be identified as key personnel with "potential high risk". In order to further pay attention to the traffic safety of minors, they can be identified as key personnel in the "personality" dimension.

[0243] Step 9: Develop a traffic safety education plan

[0244] According to the characteristics of different indicators in different dimensions of key personnel, the present invention proposes reasonable solutions in terms of the teaching methods and contents.

[0245] 9.1 Missionary Methods

[0246] Traffic safety education can be divided into two ways: online and offline. Traditional traffic safety education is mainly offline education, and traffic safety education through online new media is also developing. Different forms of education have different advantages and disadvantages and are suitable for different people. In terms of content, different specific content is suitable for people with different "personalities".

[0247] 9.2 Mission content

[0248] 9.2.1 Offline Education Content

[0249] Offline education mainly includes public welfare lectures, traffic management department education, social publicity, AR experience, etc.

[0250] 1) Public welfare lectures

[0251] The publicity and education are mainly carried out by social organizations such as community streets, schools and volunteer units, and the target groups are mainly those with low acceptance of online media and poor online education effects. Through the centralized and unified offline education of social organizations, the publicity can be in place and the expression can be sufficient to ensure that such people have basic traffic awareness. However, it is difficult to carry out such publicity and education, and the frequency is low, especially in rural areas where conditions are insufficient and it is difficult to carry out.

[0252] 2) Education by traffic control departments

[0253] After punishing traffic participants for traffic violations, we should understand their specific circumstances and conduct oral education, especially for those who have committed traffic violations many times, we should conduct one-on-one education. For key vehicles and special vehicles, we should conduct education on relevant regulations and safety awareness during routine inspections.

[0254] 3) Social publicity

[0255] Traffic control departments, communities, volunteers and other organizations set up display boards or distribute leaflets in public places to publicize the key contents of traffic regulations and traffic safety awareness. The objects and contents are relatively general and lack focus, but to a certain extent, it has raised people's attention to traffic safety.

[0256] 4) AR Experience

[0257] With the help of AR augmented reality technology, the dangerous scenarios and countermeasures that the educated people may encounter in the process of participating in traffic can be realized in a safe place. When facing the elderly and the relatively weak cultural groups, the conventional propaganda mode or text content is difficult to understand. The use of AR technology can make the educated people understand traffic regulations and improve safety awareness more intuitively and effectively.

[0258] 9.2.2 Online Education Content

[0259] Online, we mainly use text messages, news apps, public accounts, short video platforms, etc. Different media platforms are adapted to the needs of different people and different scenarios.

[0260] 1) SMS

[0261] The SMS platform can cover almost all people, and the educational effect is ideal. Compared with other media platforms, it can attract more attention and fully convey information. However, due to the limitation of words and length, the content of the education can only focus on the adjustment of road traffic laws and regulations; the information of local temporary road changes and traffic rules changes is difficult to play a role in targeted and accurate education.

[0262] 2) News APP and public accounts

[0263] News apps and public account platforms are currently the main sources of information for people in society. Web-based propaganda platforms can provide people with a place to discuss events while providing pictures, deepening people's understanding of traffic laws and regulations and strengthening traffic safety awareness. However, news platforms use text to express content, which is relatively lengthy and has little appeal to young groups, making it difficult to achieve good propaganda results.

[0264] 3) Video Platform

[0265] Videos, especially short videos, are the main force of new media and are the main source of information for young people. Short video propaganda is more compact, the content is more real, and the expression is more intuitive, which can attract more attention and deepen the impression of young people. At the same time, the background data of short video platforms is more diversified, and they can be further adjusted and optimized according to the propaganda effect after pushing the propaganda video, which has great advantages in the propaganda of key personnel. The following is a further analysis of traffic safety propaganda videos:

[0266] Through the search of keywords such as "traffic safety education", "education video", "traffic safety education video" in databases such as CNKI, Wanfang Data, and National Science and Technology Library Center, relevant literature was retrieved, and the relevant classification and grading information related to the evaluation indicators of education videos in the literature was screened and summarized, and three angles (video production, deterrence intensity, video content) and corresponding indicators and contents for evaluating the classification of traffic safety education videos were sorted out. The specific indicators and contents included in each angle are shown in Table 9-1.

[0267] Table 9-1 Specific content of video classification angles

[0268]

[0269] Traffic safety education videos are divided into three categories based on "video production", "deterrence intensity" and "video content". Class, as shown in Table 9-2.

[0270] Table 9-2 Classification of traffic safety education videos

[0271]

[0272] Step 10: Accurately match the education plan to key personnel based on the user portrait of traffic participants

[0273] Different publicity methods and contents will have different publicity effects for people of different ages, genders and educational levels. In order to rationally utilize existing resources and improve publicity effects, corresponding and appropriate publicity plans should be prepared for different types of traffic participants reflected in user portraits.

[0274] As attached Figure 8 As shown, the present invention conducts additional focused education for the key personnel screened out in step 9 based on the user portraits of the traffic participants, on the basis of the universal education for all traffic participants.

[0275] 10.1 Universal education for all traffic participants

[0276] Universal education is applicable to all traffic participants. Given the large number of target groups, it is only carried out through online education. The selection of traffic safety education methods and content is mainly based on the risk level and "personality" characteristics of the population, and combined with the "personality" characteristics of different populations.

[0277] 1) SMS platform, which conducts universal education for all traffic participants, with the content focusing on common sense and notification. Common sense education is for all traffic participants, with the content mainly on road traffic laws and regulations; notification education includes pushing information on local temporary road changes and traffic rules changes to people in different regions, and pushing information on local traffic rules when individual location changes across cities.

[0278] 2) News and public account platforms are mainly aimed at middle-aged and elderly people, suitable for people who are accustomed to reading text and traditional media. The content should be targeted and accurately pushed to specific groups. Accident case analysis articles, based on actual cases, show the actual consequences of traffic violations, and are aimed at people whose main modes of transportation and case models match; popular science articles on laws and regulations, aimed at all traffic participants, popularize basic knowledge and regulations related to road traffic safety laws; articles on exposing violations, publicly disclose serious traffic violations that have been investigated and dealt with, and enhance the deterrent effect of illegal investigations on the society; notification and announcement articles, aimed at all traffic participants, push the latest, regional changes in traffic regulations and road conditions.

[0279] 3) Short video platforms are mainly aimed at young and middle-aged people. As the main force of new media, the precise educational effect of videos should be improved from three angles: "video production", "deterrence intensity" and "targeted transportation methods".

[0280] ① The “personality” dimension of traffic participants corresponds to the “video production” perspective

[0281] In terms of video types, minors are more likely to accept cartoons and secondary editing; adults prefer original short films and picture mosaics; and the elderly have a deeper understanding of real-life and surveillance videos.

[0282] In terms of characters, minors are more accepting of short videos with celebrities and cartoons as the main characters; the elderly are more accepting of public figures such as traffic police, experts and scholars, and government officials as the main characters.

[0283] In terms of duration, minors and people with low education levels have little interest in educational videos, and are suitable for fast-paced videos of less than 30 seconds. As the main traffic participants, middle-aged people can use longer videos within their acceptance level to help them understand relevant information in more detail.

[0284] ② The “risk” dimension of traffic participants corresponds to the “deterrence intensity” angle of the video

[0285] For "high-risk" traffic participants, strong promotional videos are pushed to show the serious consequences of traffic violations and traffic accidents to play a significant warning role; for "medium-risk" or "low-risk" traffic participants, moderate and gentle promotional videos are pushed to increase people's acceptance with approachable content.

[0286] ③ The "traffic" dimension of traffic participants corresponds to the "traffic mode" angle of the video

[0287] The three indicators of "traffic" dimension, namely "travel time period", "travel OD" and "travel transportation mode", correspond to the time periods, road sections, main transportation modes and corresponding illegal behaviors or accident types mainly reflected in the video.

[0288] Based on the above analysis of the three perspectives of “video production”, “deterrence intensity” and “targeted transportation mode”, The traffic participants of each category are matched to those in Table 9-2. Such traffic safety education videos can achieve the precise education goal of video education.

[0289] 10.2 Focused education for key personnel

[0290] As attached Fig. 9 As shown, on the basis of universal education applicable to all traffic participants, precise education plans are formulated for key personnel with "high risk" in the "risk" dimension, "high potential risk" in the "traffic" dimension, and "high dangerousness" in the "personality" dimension. Among them, the education method (offline education or online education) is determined according to the "risk" dimension and the "traffic" dimension, and the education content is determined according to the "personality" dimension.

[0291] 1) Key personnel with “high risk” in the “risk” dimension

[0292] For key personnel with "high risk" in the "risk" dimension, the "online + offline" education and publicity should be combined in parallel. Such personnel have committed traffic violations many times or have committed serious traffic violations. The risk level is the highest, and the most education resources and efforts should be allocated to such personnel. Targeted education should be carried out after the punishment of their traffic violations. Such personnel have a low degree of obedience to traffic rules, and they should understand the specific situation, strengthen their traffic safety awareness, and change their mentality and driving habits through offline one-on-one education. At the same time, such personnel should be regularly given centralized and unified education to learn the content of traffic laws and regulations and traffic accident cases, understand the serious consequences and dangers of traffic violations, and thus reduce their traffic safety risks.

[0293] 2) Key personnel with “high potential risk” in the “transportation” dimension

[0294] For key personnel with "high potential risk" in the "transportation" dimension, the risk level of this type of personnel is relatively low, but the main travel time, travel mode, travel OD are dangerous, and the travel frequency is high. The potential risk source of this type of personnel is mainly the danger of the transportation mode. The focus of their publicity and education is on the education of laws and regulations related to special vehicles and safety awareness. Offline education should be carried out during the annual driving qualification review to improve safety awareness and enhance the effectiveness of publicity and education.

[0295] 3) Key personnel with “high risk” in the “personality” dimension

[0296] For key personnel with "high risk" in the "personality" dimension, offline education should be carried out by communities, schools and other social organizations for the elderly and minors. The elderly generally have a low acceptance of online media, and the implementation of education and publicity needs to be promoted through traditional offline education. Minors need to increase their attention to safety awareness through school platforms and further supervise the formation and improvement of their habits. Groups with low cultural levels and minors often have a strong curiosity and desire to explore new things. Through new forms of education such as AR augmented reality technology, it is possible to achieve dangerous scenarios and countermeasures that may be encountered in the process of participating in traffic in a safe venue for the educated population. Through interactive methods, the educated population can understand the dangers of traffic violations and accidents more intuitively and effectively, so as to have a deeper understanding of the importance of traffic safety, greatly improve their acceptance of traffic safety knowledge, and enhance traffic safety awareness.

[0297] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solution of the present invention can be modified or replaced by equivalents without departing from the spirit and scope of the technical solution of the present invention.

Claims

1. A precise traffic safety education system based on traffic participant user portraits, characterized in that: include: Traffic participant multi-dimensional data processing module, which is used to construct indicators of the three dimensions of "personality", "risk" and "traffic" based on the basic personal information of traffic participants; A personality category classification module is used to classify the "personality" dimension indicators of traffic participants into multiple personality categories using a chameleon clustering algorithm, and to classify the danger level according to the personality categories; The risk level classification module is used to classify traffic participants into corresponding risk levels according to their "risk" dimension indicators; A potential risk level classification module is used to classify traffic participants into corresponding potential risk levels according to their "traffic" dimension indicators; Traffic participant user portrait module, which is used to classify traffic participants into multiple types of user portraits according to their personality categories, risk levels, and potential risk levels; The education program matching module is used to match education programs for different types of user profiles; In the multi-dimensional data processing module for traffic participants, indicators of three dimensions of "personality", "risk" and "traffic" are constructed based on the basic personal information of traffic participants, including: "Personality" dimension indicators: gender, age, education, occupation; "Risk" dimension indicators: traffic violation information, traffic accident information; "Transportation" dimension indicators: travel time, travel mode, travel OD, travel frequency; The risk level classification module includes: Traffic violation assignment unit, which is used to assign a value to the traffic violation according to the traffic violation information in the "risk" dimension indicator of the traffic participant. The value is determined by the traffic participant's motor vehicle violation score, non-motor vehicle violation score, and other violation score; A traffic accident situation assignment unit, which is used to assign a value to the traffic accident situation according to the traffic accident information in the "risk" dimension indicator of the traffic participants. The value is determined by the number of traffic accidents, the corresponding level of the traffic accidents, and the points assigned to the responsibility; A risk assignment and classification unit, which is used to calculate the sum of the traffic violation assignment and the traffic accident assignment to obtain a risk value, and classify the risk value into "low risk", "medium risk" or "high risk"; The potential risk level classification module includes: The weight assignment unit is used to determine the weights of the four "traffic" dimension indicators of the traffic participants, namely, travel time, travel mode, travel OD, and travel frequency, by using the hierarchical analysis method, and mark the four indicators as four evaluation labels; The potential risk value assignment and grade division unit is used to calculate the product of each evaluation label value of each traffic participant and the weight of the evaluation label and accumulate them to obtain the potential risk value, and divide the potential risk value into "low risk", "medium risk" or "high risk".

2. The traffic safety precision education system based on traffic participant user portrait according to claim 1 is characterized in that: The basic personal information of traffic participants includes gender, age, education, occupation, traffic violation information, traffic accident information, traffic monitoring data, traffic checkpoint data, traffic behavior mobile phone signaling, and traffic payment information.

3. The traffic safety precision education system based on traffic participant user portrait according to claim 2 is characterized in that: If the basic personal information of traffic participants is missing, the missing values ​​will be filled in according to the participants with the highest duplication rate in the database.

4. The traffic safety precision education system based on traffic participant user portrait according to claim 1 is characterized in that: The travel OD includes travel between cities, travel between rural areas, and travel between cities and rural areas; the "traffic" dimension indicators are obtained based on traffic monitoring data, traffic checkpoint data, traffic behavior mobile phone signals, and traffic payment information.

5. The traffic safety precision education system based on traffic participant user portrait according to claim 1 is characterized in that: In the personality category division module, the chameleon clustering algorithm is used to divide the "personality" dimension index into multiple personality categories, including: The "personality" dimension index of each traffic participant is numerically processed to obtain the characteristic vector of the "personality" dimension index; The characteristic vector of the "personality" dimension indicator of each traffic participant is taken as a node, and the Euclidean distance between each node and all other nodes is calculated; Take the K nodes with the closest Euclidean distance as the nearest neighbor points and build edges, take the inverse of the Euclidean distance as the edge weight, and construct a KNN graph; Treat the KNN graph as a cluster, divide it into unconnected subclusters, and put them into the list to be merged; For each subcluster in the list to be merged, calculate its similarity with other subclusters. If the maximum similarity exceeds the threshold, merge the subcluster with the subcluster with the maximum similarity. Traverse all sub-clusters in the list to be merged and complete the merger; each sub-cluster after the merger is regarded as a personality category; The personality categories were named according to the most repeated indicator of traffic participants within each cluster after merging, and were classified as "high-risk" or "low-risk".

6. The traffic safety precision education system based on traffic participant user portrait according to claim 1 is characterized in that: In the traffic participant user portrait module, the chameleon clustering algorithm is used to divide the personality categories, risk levels, and potential risk levels of traffic participants into multiple types of user portraits, including: The personality category, risk level, and potential risk level of each traffic participant are used as feature vectors to calculate the Euclidean distance between each node and all other nodes; Take the K nodes with the closest Euclidean distance as the nearest neighbor points and build edges, take the inverse of the Euclidean distance as the edge weight, and construct a KNN graph; Treat the KNN graph as a cluster, divide it into unconnected subclusters, and put them into the list to be merged; For each subcluster in the list to be merged, calculate its similarity with other subclusters. If the maximum similarity exceeds the threshold, merge the subcluster with the subcluster with the maximum similarity. Traverse all sub-clusters in the list to be merged and complete the merger; each sub-cluster after the merger is regarded as a user profile; The user profiles are named according to the most repeated features of the traffic participants in each cluster after merging.

7. The traffic safety precision education system based on traffic participant user portrait according to claim 1 is characterized in that: The missionary program matching module includes: Key Personnel Identification Unit, which is used to identify key personnel in the three dimensions of "personality", "risk" and "traffic" according to the risk level, potential risk level and personality category of traffic participants; The education program customization unit is used to customize universal education programs for various types of user portraits from three perspectives: video production of traffic safety education videos, deterrence intensity, and traffic modes; and to customize key education programs for key personnel in different dimensions.