基于进化计算的交叉口交通流聚类优化关键技术研究

发布时间：2018-01-05 13:04

本文关键词：基于进化计算的交叉口交通流聚类优化关键技术研究　出处：《南京邮电大学》2016年硕士论文　论文类型：学位论文

【摘要】：交叉口交通路段是当前交通拥堵和事故发生的聚集地,聚类分析常常用于对这些交通路段进行规划和分类,但目前实际采用的聚类算法通常存在对聚类边界点的识别不够准确、对初始聚类中心的选取具有随机性等问题。为此,本文利用进化计算来优化形成交叉口交通流聚类方案,解决上述问题,以提高交叉口交通流聚类分析性能。本文建立面向交叉口交通流聚类的进化计算模型,在此基础上,选取遗传算法和粒子群算法分别对交叉口交通流K-means聚类和密度聚类进行优化,主要研究工作如下:(1)提出一种基于遗传算法的交叉口交通流K-means聚类优化算法,对数据复杂的交叉口交通流进行聚类。该算法首先将随机选取出来的初始聚类中心作为第一代种群进行遗传操作,将初始种群进行二进制编码;然后在初始种群中选择染色体进行交叉、变异操作,由改进的交叉算子进行下一代种群的选取;最后根据交叉口交通流聚类特征给出评价函数,通过该评价函数选择最优解。实验表明,该算法能够快速地对交叉口交通流进行聚类,有效降低初始聚类选取时所产生的不必要损耗,提高交叉口交通流聚类的精确度。(2)提出一种基于粒子群算法的交叉口交通流密度聚类优化算法,解决交叉口大数据量交通流聚类问题,首先结合改进的局部异常因子(LOF)算法将差别较大的交叉口进行排除;然后根据初始参数和MinPts进行核心对象的筛选,将核心对象的核心距离由高到低进行排序,选取初始聚类中心;最后根据粒子群算法对初始数据进行编码,采用粒子群算法的速度和位置公式来更新种群。实验表明,该算法能够对交叉口大数据量的交通流进行聚类,降低初始交叉口交通流聚类中心选取的错误率,更好地对差异性较大的交叉口进行识别。
[Abstract]:Intersection traffic section is the gathering place of traffic congestion and accident at present. Cluster analysis is often used to plan and classify these traffic sections. However, the existing clustering algorithms usually have some problems, such as the identification of the clustering boundary points is not accurate enough, and the selection of the initial clustering center is random. In order to improve the performance of intersection traffic flow clustering analysis, this paper uses evolutionary computation to optimize the traffic flow clustering scheme of intersection to solve the above problems. In this paper, an evolutionary computing model for intersection traffic flow clustering is established. On this basis, genetic algorithm and particle swarm optimization algorithm are selected to optimize K-means clustering and density clustering of traffic flow at intersection. The main research work is as follows: (1) A K-means clustering algorithm for intersection traffic flow based on genetic algorithm is proposed. Firstly, the initial cluster center selected by random is used as the first generation population for genetic operation, and the initial population is coded by binary. Then the chromosomes are selected for crossover and mutation operation in the initial population, and the next generation population is selected by the improved crossover operator. Finally, the evaluation function is given according to the intersection traffic flow clustering characteristics, and the optimal solution is selected by the evaluation function. The experiment shows that the algorithm can cluster the intersection traffic flow quickly. In order to improve the accuracy of intersection traffic flow clustering, an optimization algorithm for intersection traffic flow density clustering based on particle swarm optimization (PSO) is proposed. In order to solve the problem of traffic flow clustering with large amount of data at intersections, the improved local anomaly factor (LOF) algorithm is used to eliminate the intersections with large differences. Then the core objects are filtered according to the initial parameters and MinPts, the core distance of the core objects is sorted from high to low, and the initial clustering center is selected. Finally, the initial data is coded according to particle swarm optimization algorithm, and the speed and position formula of particle swarm optimization algorithm is used to update the population. The experiment shows that the algorithm can cluster the traffic flow with large amount of data at intersection. The error rate of selecting the traffic flow clustering center at the initial intersection is reduced, and the intersection with greater difference is better identified.
【学位授予单位】：南京邮电大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：U491;TP18

【相似文献】