时空亚频繁co-location模式挖掘  

Spatiotemporal Sub-prevalent Co-location Pattern Mining

作　　者：李新源 陈红梅[1] 肖清[1] 王丽珍[1] LI Xin-yuan;CHEN Hong-mei;XIAOQing;WANG Li-zhen(School of Information Science and Engineering,Yunnan University,Kunming 650504,China)

机构地区：[1]云南大学信息学院,昆明650504

出　　处：《西南大学学报（自然科学版）》2020年第11期68-76,共9页Journal of Southwest University(Natural Science Edition)

基　　金：国家自然科学基金项目(61662086,61966036);云南省创新团队项目(2018HC019).

摘　　要：空间co-location模式挖掘是空间数据挖掘的重要分支,在环境保护、公共交通、位置服务和城市计算等领域得到广泛应用.与基于团实例模型的传统模式相比,基于星型实例模型的空间亚频繁co-location模式可以揭示空间特征更丰富的空间关系.然而,现有空间亚频繁模式没有考虑空间数据的时间特性,而时间却是空间数据的重要维度.因此,该研究考虑空间实例的位置时变性,基于星型实例模型的时空亚频繁co-location模式进行挖掘.首先,提出了时空亚频繁co-location模式及其度量指标:时间亚频繁度;其次,证明了时间亚频繁度的反单调性(向下闭合性),提出了有效的时空亚频繁模式挖掘算法;最后,通过大量实验,验证了所提算法的有效性及时空亚频繁模式的实用性.Spatial co-location pattern mining is an important research area,and has been widely applied in various fields such as environment protection,public transport,location-based services and urban computing.Compared with the traditional pattern based on clique instances,the spatial sub-prevalent pattern based on star instances can reveal richer spatial correlations among features.However,the temporal characteristic of spatial data,which is an important dimension of spatial data,is not considered by the current spatial sub-prevalent pattern.Hence,in this paper,a spatiotemporal sub-prevalent pattern based on star instances is presented by analyzing spatial instances whose locations change over time.Firstly,a metric to measure the pattern called“time sub-frequency”is proposed.Then,the anti-monotonicity of the metric is proven,and an efficient algorithm for mining the pattern is designed.Finally,the validity of the algorithm presented herein and the applicability of the pattern are verified through a number of experiments.

关 键 词：空间数据挖掘 时空数据 时空亚频繁co-location模式 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]