| Since the reform and opening up,great changes have taken place in the structure of China’s bus industry,which not only improves the convenience of people to go out,but also makes the traffic safety problem increasingly serious,and the public transport safety standards need to be improved.Among them,the problem of bus speeding has always been a difficult problem in the field of traffic safety.With the maturity of data mining algorithm and the support of onboard Internet of things data,analyzing the information of speeding section in bus on-board data provides a new means and direction for strengthening road supervision.In this thesis,through the on-board data obtained by the intelligent terminal of the bus,the information of the road section with frequent overspeed is obtained by using the DBSCAN algorithm.The specific research work is as follows:(1)Analyze the vehicle-mounted data set,determine the clustering methodology,perform data preprocessing on the vehicle-mounted data set,including exploratory data analysis and data cleaning,and combine the characteristics of the clustering algorithm to select the DBSCAN algorithm suitable for processing the data set to complete the speeding road section Clustering.And according to the structure of the data itself,combined with the contour coefficient theory,adaptively determine the neighborhood radius and the density threshold to improve the clustering accuracy.(2)Optimizing the memory requirement of traditional DBSCAN algorithm in finding the neighborhood of core points,and a method based on grid cell division and minimum neighborhood query is proposed.By searching for the critical region and the minimum query region,a large number of queries and calculation operations are reduced,and the DBSCAN algorithm is effectively shortened.The running time when processing big data,the clustering speed is improved.(3)To study the influence of speed weighting factors on speeding sections,and combining weighting factors with DBSCAN algorithm to analyze speeding sections.According to the severity of speeding,speeding sections can be divided into severe speeding areas,general speeding areas,and slight speeding areas,providing relevant traffic prevention and control Targeted and forward-looking decision-making ideas.The experimental results based on DBSCAN algorithm show that the improved DBSCAN algorithm is faster than the traditional algorithm when dealing with the same size of vehicle data sets.At the same time,the clustering research based on the speed weight factor can divide the speeding sections according to the severity of speeding,which makes the analysis results of speeding sections more precise and accurate. |
