Research On Key Technologies Of Intersection Traffic Flow Clustering Optimization Based On Evolutionary Algorithm

Intersection traffic road is a gathering place of traffic congestion and accidents. Cluster analysis is used to classify these traffic flow sections, but at present, the practically used clustering algorithms usually have many problems, such as not enough accurately identifications of the clustering boundary points, randomly selections of the initial clustering centers. Therefore, this thesis uses evolutionary computation to optimize the intersection traffic flow clustering, solves the above problems, and improves the performance of clustering analysis on the intersection traffic flow.This thesis establishes an evolutionary computation model of intersection traffic flow clustering, and based on the evolutionary computation model, selects genetic algorithm and particle swarm optimization algorithm to optimize the K-means clustering and density clustering of the intersection traffic flow, respectively. The main research work is as follows:(1) This thesis presents a K-means clustering optimization algorithm for intersection traffic flow based on genetic algorithm, which clusters the intersection traffic flow with complex data. Firstly, the algorithm selects initial cluster centers as the first population for genetic manipulation. The initial population is encoded by binary encoding. Then the algorithm chooses chromosomes to carry out the crossover and mutation operation from the first population. The next population is produced by the improved crossover operator. Finally, according to the clustering features of the intersection traffic flow, the algorithm proposes the evaluation function, which is used to choose the optimum solutions. The experiments show that the proposed algorithm can be used in intersection traffic flow clustering optimization, effectively reduce the unnecessary loss of the initial cluster centers’ selection, and improve the accuracy of the intersection traffic flow clustering.(2) This thesis presents a density clustering optimization algorithm for intersection traffic flow based on particle swarm optimization algorithm, and solves the clustering problem of intersection traffic flow with large data. Firstly, the algorithm uses the local outlier factor(LOF) algorithm to exclude the intersections which have great difference. Then the algorithm screens the core objects by the initial parameters of and MinPts. The core distances are ranked high to low to select the initial cluster centers. Finally, the initial data are encoded by the particle swarm optimization algorithm. The populations are updated by the velocity and position formula of the particle swarm optimization algorithm. The experimental results show that the algorithm can cluster the intersection traffic flow with large data, reduce the selection error rate of the initial intersection traffic flow clustering centers, and better identify the intersections with large difference.