基于方差优化谱聚类的热点区域挖掘算法

Hot Region Mining Algorithm based on Variance Optimization Spectrum Clustering

关键词:  K-medoids算法  谱聚类  热点区域  停留点  交通拥堵

Key words:  K-medoids algorithm  spectral clustering  hot region  stop point  traffic congestion

摘要:

为改善交通拥堵的情况，本文利用聚类分析方法对移动轨迹数据进行挖掘，识别居民出行的热点区域。传统的Ng-Jordan-Weiss （NJW）谱聚类算法常使用K-means聚类算法来实现最后的聚类操作，然而K-means聚类算法存在对初始值敏感、容易陷入局部最优的缺陷，影响对热点区域的挖掘结果。因此，本研究将方差优化初始中心的K-medoids聚类算法运用到谱聚类算法最后聚类阶段，提出基于方差优化谱聚类的热点区域挖掘算法（Hot Region Mining algorithm based on improved K-medoids Spectral Clustering，HRM-KSC），然后在真实的轨迹数据集上进行试验。试验结果发现，HRM-KSC算法聚类结果的轮廓系数更高，表明HRM-KSC算法改善了NJW谱聚类算法，提高了聚类质量。

Abstract:

In order to improve the traffic congestion,this article uses the cluster analysis approach to mine the mobile trajectory data and identify the hot region of residents' travel. The traditional Ng-Jordan-Weiss (NJW) spectral clustering algorithm often uses K-means clustering algorithm to achieve the final clustering operation. However,K-means clustering algorithm has the disadvantages of being sensitive to the initial value and easy to fall into the local optimum,which will affect the mining results of hotspot area. Therefore,the K-medoids clustering algorithm of variance optimization initial center is applied to the final clustering stage of the spectral clustering algorithm,and a Hot Region Mining algorithm based on improved K-medoids Spectral Clustering (HRM-KSC) is proposed,and then experiment on real trajectory data sets. The experiment results find that the HRM-KSC algorithm clustering results have higher silhouette coefficient,which indicates that the HRM-KSC algorithm improves the NJW spectral clustering algorithm and the clustering quality.