点集数据不规则形状时空异常聚类模式挖掘研究

Research on Irregularly Shaped Spatio-Temporal Abnormal Cluster Pattern Mining for Spatial Point Data Sets

传统扫描统计方法在进行时空异常聚类模式挖掘时,受扫描窗口形状的限制,不能准确地获取聚类区域形状。提出一种改进的不规则形状时空异常聚类模式挖掘方法stAntScan。新方法基于26方位时空邻近单元格构建时空邻接矩阵,再对蚁群最优化扫描统计方法进行改进,使其能适应三维大数据量的时空区域扫描。模拟数据和真实微博签到数据的实验证明,stAntScan能有效地识别时空范围内的不规则形状异常聚类,并且准确性较经典的SaTScan方法高。

Spatio-temporal abnormal cluster pattern is an important spatial point pattern.The pattern results can reflect the distribution and evolution of spatio-temporal events timely and accurately.Early researches has verified the scan statistic based clustering methods are very effective in detection spatial and spatio-temporal abnormal cluster pattern.However,due to the fixed shape of scan window,traditional scan statistic based clustering methods have limitation on obtaining exact shape and size of cluster.This paper proposed an improved irregularly shaped spatio-temporal abnormal cluster pattern mining algorithm stAntScan.The algorithm constructs the spatio-temporal neighborhood matrix by a newly defined 26 directions spatio-temporal neighbor cells.Then the algorithm improves the ant colony optimization based method to fit for spatio-temporal scanning on three-dimensional large data set.In the end,the Monte Carlo simulation method is used to test the significance of clusters.Experimental results on both simulated data and real Weibo check-in data have testified the efficiency and accuracy of stAntScan on irregularly shaped spatio-temporal abnormal cluster pattern mining.And compared with the classical SaTScan,it gets much better results in finding exact shape and size of clusters.