Research On Data Mining Algorithm Of Spatio-temporal Data

With the development of mobile Internet technology and data collection technology,a large number of Spatio-temporal data including time,space and other information have emerged.How to mine potentially useful knowledge from these Spatio-temporal data and use it to guide social services such as decision-making and business analysis has become a new research problem and hot spot in the field of data mining research.This paper studies the problem of Spatio-temporal data mining,which mainly includes the following two aspects.(1)Research on the method of Region of Interest discovery under Spatio-temporal data.Aiming at the problem that the existing spatial clustering method has a large time cost and is not suitable for the characteristics of Spatio-temporal data,a Spatio-temporal Region of Interest discovery algorithm SPROI is proposed,which is a clustering algorithm based on grid and Spatio-temporal density.On the one hand,the Spatio-temporal density is given,that is,the weighted sum of access frequency and dwell time,fully considering the time and space constraints;on the other hand,the use of grid-based clustering methods improves the efficiency of the algorithm when multiple and repeated clustering is required.The main process of the SPROI includes;first,divide the spatial range into cells,calculate the Spatio-temporal density of the cell where the data is located,and find the core cell that meets the density threshold.Then merge and fiter the core cells,and finally get the Spatio-temporal Region of Interest.Experiments show that the algorithm is feasible and efficient.(2)Research on frequent pattern mining methods under Spatio-temporal data.Due to the particularity of Spatio-temporal data,the existing frequent pattern mining algorithms are not suitable for the mining problem under Spatio-temporal data.A frequent pattern mining algorithm SPMQT-mining for Spatio-temporal data is proposed in this paper.The algorithm considers the dynamics and sequence of Spatio-temporal data,uses a sliding window model,proposed a summary structure SPMQT to compress and store Spatio-temporal data,and mines Spatio-temporal frequent sequence patterns based on Prefix Span method.The main process of the algorithm includes:first,the Spatio-temporal data is converted into a sequence of Regions of Interest through the SPROI algorithm.Then,according to the characteristics of Spatio-temporal data,a summary structure SPMQT including matrix,prefix tree,and queue is constructed.SPMQT is used to compress and store Spatio-temporal data,while separating the support information,sequence information and index information required for mining in Spatio-temporal data.Finally,the projection database is generated recursively based on the Prefix Span algorithm to complete mining.Experiments show that the algorithm has good space-time efficiency.