轨迹聚类, 城市, 热点区域, 网格密度, 峰值挖掘算法

Urban pedestrian flow is time-dependent, and the static density peak is difficult to reflect the dynamic change pattern. By identifying highp-density areas in different time periods through temporal clustering, the spatio-temporal evolution laws can be captured. For this purpose, a peak mining algorithm for grid density in urban hotspot areas based on trajectory clustering is proposed. The trajectory distance is combined to obtain the regional dynamic group information, the position points are arranged and combined according to the trajectory clustering timestamp, the trajectory length is adjusted, and abnormal data is cleaned. The comprehensive state of the peak data of grid density in urban hotspot areas is considered for multi-event adjustment, matching technology is used to obtain the location point road sections, the similarity between data is calculated, the center points are allocated according to the clustering state, and the comprehensive spatial attributes of the trajectory points are determined. The center point of the grid cell is selected as a representative to construct a peak mining model for grid density in hot areas, the minimum distance of the mined data is calculated, peak mining labels are generated, and the peak mining of grid density in urban hot areas is completed. The case analysis shows that after the application of the proposed algorithm, the DBI ( Davies-Bouldin Index) under different numbers of clustering centers is relatively small, close to 0. This proves that the algorithm has compact clustering within clusters, high separation degree, good clustering effect and high-quality robustness after clustering.

trajectory clustering, city, hotspot area, grid density, peak mining algorithm