Research And Application Of Parallel FP-Growth Mining Algorithm For Freeway Traffic Crash Data Under Load Balancing Constraints

The number of deaths caused by freeway traffic accidents is increasing year by year.Deeply digging and analyzing the potential correlations between the attributes of freeway traffic accidents and revealing the coupling mechanism will be an effective means to explore the causes of freeway traffic accidents.However,the existing research has certain limitations in the design and application of association rule mining algorithms: the association rule mining algorithm of ordinary serial mode usually generates a large number of candidate item sets during the mining process,which requires a large amount of running memory to support,and it cannot Effectively mining massive traffic accident data;at the same time,most studies only focus on association rules with high support and confidence levels,and it is easy to ignore potential important traffic accident risk factors;in addition,common parallel association rule mining algorithms generally have unbalanced load The problem,which in turn leads to low computational efficiency.Based on this,this research builds a multi-dimensional and multi-layer accident cause analysis system.Under the premise of considering load balancing,the existing FP-Growth algorithm is optimized in parallel based on the Hadoop platform to achieve massive freeway traffic accident data Efficient and accurate association rule mining and calculation;further,based on the mining results of the coupling mechanism between the various influencing factors of freeway traffic accidents,in-depth analysis and revealing the causes of freeway traffic accidents,and providing guidance for accident risk prevention and control.The main research work of this paper is as follows:(1)Freeway traffic accident data collection and sample structure design.Preprocessed the acquired historical accident data set of Washington State,USA,and obtained 29 influencing factors in six dimensions of human(driver),vehicle,road,environment,time and accident,and constructed a six-dimensional three-layer traffic accident Analyze the model and realize the sample building of multi-dimensional database.Based on the characteristics of mathematical statistics,the distribution rules and characteristics of traffic accidents are analyzed from the four dimensions of human(driver),vehicle,road,and environment.(2)Association rule mining algorithm and Hadoop platform technology related theory combing.Combined with the characteristics of the freeway traffic accident data set,the related technical methods of data mining are summarized,the concepts of association rule mining,evaluation indicators and related classic algorithms are sorted out,and the basic principles and application implementation methods of the Hadoop platform are further introduced for follow-up research.The expansion laid the theoretical foundation.(3)The design and implementation of parallel optimized FP-Growth algorithm considering load balancing.For freeway traffic accident data,a parallel FP-Growth algorithm considering load balancing is proposed to conduct multi-dimensional and multi-layered massive accident data mining: First,the overall architecture of the parallel FP-Growth algorithm is designed and combined with related theories.Code implementation work;further,for the algorithm’s operating efficiency,load balancing algorithm is applied to optimize the grouping in the parallel FP-Growth algorithm.The results show that the optimized FP-Growth association rule mining algorithm can distribute the load more evenly on the Hadoop cluster nodes on.The actual test results of the algorithm show that the parallel FP-Growth algorithm under load balancing constraints runs faster when processing large data sets.This improved algorithm makes full use of the Hadoop cluster resources,while retaining the original advantages of the association rule mining algorithm.It is more suitable for mining large data sets of traffic accidents.(4)The application of improved FP-Growth algorithm in the cause analysis and risk prevention and control of freeway traffic accidents.Based on the case study of freeway traffic accident data in Washington State,USA,when the model is constructed through the improved FP-Growth algorithm proposed in this paper,the association rule mining model for freeway traffic accident data is further improved from the perspective of multi-dimensional interaction.That is,mining association rules from the three perspectives of accident dimension autocorrelation,single dimension and accident dimension,and accident multi-dimensional interaction,extracts the association rules with high support and high confidence in the mining results,and conducts traffic accidents based on the mining results Coupling mechanism analysis.Extract the association rules with high support and confidence in the mining results for interpretation,and further analyze and explain the internal mechanism of freeway traffic accidents.Furthermore,from the perspectives of high support,high confidence,and low support,countermeasures and suggestions are put forward to control the risk of freeway traffic accidents and improve the level of freeway traffic safety.