Research On Traffic Accident Factors Mining And Prediction Model Based On Mean Clustering

In recent years,with the rapid development of social economy,the road transportation industry as a basic industry is growing,the investment in infrastructure transportation is increasing,and the construction speed of expressways,ordinary roads and urban roads is accelerating.At the same time,the rate of road traffic accidents is gradually rising,and traffic safety has become an urgent problem to be solved.Road traffic accidents have the characteristics of complexity and diversity.The causes of accidents are often multifaceted,which are directly or indirectly related to many factors such as people,vehicles,environment and roads.It is not rigorous to attribute all the causes of accidents to the personal responsibility of drivers.With the rise of the era of big data,data mining technology and machine learning methods began to be applied to traffic accident analysis.Making full use of accident data to explore the hidden laws behind accidents and establish a scientific accident prediction model is of great significance to improve traffic safety.The main work of this paper is as follows:1.In the mining of traffic accident factors,aiming at the low efficiency of Apriori algorithm in the face of large data sets,this paper starts with the number of iterations of the algorithm to search the database,and locally optimizes the Apriori algorithm by adding the candidate itemset counter.The traditional Apriori algorithm needs to traverse the database many times to filter the frequent itemsets in the pruning stage,but it can determine the frequent itemsets only by traversing the database once.By designing simulation experiments and comparing with a variety of association rule algorithms,the experimental results show that the optimized Apriori algorithm has higher mining efficiency in the case of low support threshold and large amount of data.Aiming at the problem of weak clarity of association results in traffic accident oriented association analysis,using the good clustering performance of clustering algorithm,K-means clustering algorithm and optimized Apriori algorithm are combined in series.Firstly,Kmeans clustering algorithm is used to classify and cluster the original accident data,and three accident levels are divided into minor accident,serious accident and major accident,Then the optimized Apriori algorithm is used to analyze the association of each level of accidents,and the strong association rules in each level of accidents are mined.Through experiments,compared with the original association rule algorithm,the association rule algorithm based on clustering can mine the relationship between accident severity and various factors more clearly and intuitively.2.In the study of accidents prediction model,genetic algorithm(GA)is used to improve the defects of traditional BP neural network and build GA-BP model.Because of the randomness of the initial weight and threshold in the traditional BP neural network,it is easy to fall into the local optimal solution.According to the reading literature,this paper finds that genetic algorithm has excellent global optimization ability.Genetic algorithm is used to iteratively optimize the initial parameters of BP neural network,and the optimal weight and threshold are calculated and input into BP neural network.Simulation experiments verify the effectiveness of GA optimization,which establishes the foundation for the construction of the following combination model.In the process of experiments,this paper found that in the prediction results of a single model,the GABP model has a higher accuracy rate for negative samples,while the support vector machine(SVM)model has a higher accuracy rate for positive samples.Therefore,in this paper,the GA-BP model and the SVM model are combined in parallel to build a traffic accident prediction model,so that the advantages of the two are complementary,and the prediction results are output by assigning different weights to the two models.The experimental results show that the combined model has better performance and higher accuracy than the single models.