Analysis And Research Of Driving Behavior Based On Clustering And Association Rules

In this paper,the driving behavior characteristics of association rules mining,to improve the K-means algorithm and improved Apriori algorithm as the main research goal,and use the standard UCI datasets and Teddy cup data mining contest driving behavior datasets to verify the effectiveness of the algorithm.Firstly,the driving behavior data is preprocessed,and the abnormal data is cleaned and deleted to prevent the influence of dirty data on the mining results.Since Apriori algorithm can only mine association rules between discrete data in essence,and driving behavior characteristic parameters are continuous in space,the K-means algorithm is used to cluster driving behavior data,and the partition results will be used as the basis for subsequent discretization of driving behavior data.Using the traditional K-means algorithm to cluster the driving behavior characteristic parameters directly will lead to unstable clustering results.In order to solve this problem,an improved k-means algorithm is proposed to select the initial clustering center based on the decision value.Based on local density and relative distance,the algorithm constructs two-dimensional decision value graph of samples;the algorithm can automatically select the data object with the largest decision value as the initial center point,which ensures the stability of clustering.Experimental results show that the improved algorithm is more stable than k-means algorithm,the number of iterations is reduced,and the clustering performance is excellent on real datasets.Finally,the improved K-means algorithm is used to cluster the driving behavior segments.According to the contour coefficient,the optimal number of clusters is 3.Therefore,the driving behavior is divided into three types:dangerous type,irritable type and stability type,in which dangerous type accounts for18.35%,irritable type accounts for 25.65% and robustness accounts for 56%.Secondly,in order to mine the implicit relationship between driving behavior characteristic parameters,this chapter proposes an Apriori algorithm based on Boolean matrix reduction,which is transformed into Boolean matrix by scanning the transaction database.In the process of generating frequent itemsets,the reduction matrix is constrained according to the minimum support.Finally,the performance of the proposed algorithm is verified by using UCI standard data set and driving behavior datasets.The experimental results show that the improved algorithm is better than the traditional Apriori algorithm in running time and memory consumption.Finally,the Apriori algorithm based on Boolean matrix reduction is used to mine the association characteristics of driving behavior,and it is concluded that dangerous driving behaviors mainly exist many bad driving habits,such as speeding,fatigue driving,emergency braking and so on.The algorithm based on clustering and association rules proposed in this paper can accurately mine the association relationship between driving behaviors.It is proved that the improved algorithm is valuable in the analysis and decision-making of driving behaviors in theory and experiment.