Research On Analysis And Prediction Of Illegal Driving Behavior Of Truck Drivers Driven By Big Dat

With the rapid development of social economy,especially in recent years,the number of motor vehicles in our country is increasing.The rapid increase in the number of motor vehicles brings hidden dangers to the safety of road traffic.On the one hand,the occurrence of lorry drivers’ illegal driving behaviors will increase the pressure of traffic safety in our country,which causes irreparable loss of life and property;On the other hand,it will reduce the efficiency of road freight transportation and hinder the healthy development of our logistics industry.As the focus of road traffic safety research in recent years,the problem of preventing road safety accidents and improving traffic safety has made a new breakthrough with the development of big data information intelligence,and how to accurately and effectively use and analyze the natural driving data of truck drivers has become a key technical issue to avoid traffic safety accidents.Therefore,using data mining technology to mine and analyze the data of truck drivers can not only prevent the potential hazards of highway safety brought by the dangerous driving behavior of truck drivers,but also reduce the losses of freight platforms,improve the efficiency of highway freight transport,and promote the healthy development of the logistics industry.Based on this problem and the natural driving data of truck drivers recorded by on-board devices,this paper uses deep learning technology to carry out data mining on big data.The main research contents are as follows:(1)Prediction of dangerous Driving Behavior of truck drivers based on Data mining.Aiming at the problems of large data volume,high dimension,strong timing and difficult feature extraction of truck driving information,this paper proposes an improved self-attention mechanism based dangerous driving behavior prediction method for truck drivers.First,the extreme gradient lifting method was used to conduct feature correlation analysis,and then the features highly correlated with dangerous driving behavior were selected.The selected features were interfaced with the convolutional neural network and the long and short term memory network for model training.In order to solve the nonlinear fitting ability of the model,the Gaussian error linear unit was introduced in this paper to improve the self-attention mechanism module.Experimental results show that the improved model method in this paper significantly improves the feature fitting ability and improves the prediction accuracy.(2)Research on Truck driver’s Driving style Recognition based on Integrated learning.Aiming at the problems of low stability and low accuracy caused by the single model of traditional truck driver’s driving style recognition algorithm.In this paper,an integrated learning approach based on truck driver driving risk level recognition is studied.Firstly,the truck drivers are divided into normal type,low risk type and high risk type by elbow method,and the risk types of truck drivers are clustered respectively by K-means++ and Fuzzy C-means algorithm(FCM).The two clustering methods were divided into data in the same cluster as "labeled data" and data in different clusters as "unlabeled data".Then,the " labeled data" was divided into training sets and test sets and input into the classification algorithm of subsequent integration learning.support vector machines were used,Classification And Regression Tree(CART)and BP neural network are used as weak classifiers,and the MV-SCB model is constructed to classify "unlabeled data" by using the combination strategy of Majority Vote.To achieve a complete classification of truck drivers’ driving risk levels.The experimental results show that compared with the traditional clustering algorithm and the single clustering and classification algorithm,the proposed model not only has strong stability in the case of a large amount of data,but also has further improved accuracy.