As one of the critical technologies of geo-spatial data mining,spatio-temporal anomaly detection has the capacity of providing key breakthroughs for deeply revealing the evolution mechanism of geographic processes.Promoted by the big data and artificial intelligence technology,the transformation from data-driven to knowledge-driven modeling is the development tendency for the intelligent detection of spatio-temporal anomalies from geographic big data.This paper systematically sorts out the development process and the mainstream study ideas of current spatio-temporal anomaly detection.Through analyzing the dialectical relationships among data,information and knowledge,a unified description framework of spatio-temporal knowledge is constructed by integrating geographic variables,space basis,spatio-temporal relationships and knowledge types.Then,the connotation of bidirectional driving between spatio-temporal knowledge and spatio-temporal anomalies is elaborated with the help of practical cases.The implementation path for intelligent detection of spatio-temporal anomalies is further proposed,which includes spatio-temporal knowledge correlation modeling,spatio-temporal anomaly intelligent detection and spatio-tem-poral anomaly-based knowledge dynamic updating,so as to support both the reliable spatio-temporal anomaly detection and the credible spatio-temporal knowledge services.
关键词
时空异常/地理大数据/时空知识/知识图谱/深度学习
Key words
spatio-temporal anomaly/geographical big data/spatio-temporal knowledge/knowledge graph/deep learning
维普
万方数据