Co-occurrence Pattern Mining And Analysis Based On Spatial-temporal Trajectory Data

Co-occurrence patterns in urban scenarios is a newly-proposed special human mobility which focuses on urban regional relationships,and it presents high application value in various fields,such as urban planning and business intelligence.However,the time dependence and high-order nature hidden in the complex sequence transition law and multi-level predictability of spatial-temporal co-occurrence patterns,and the heterogeneity and sparsity of spatiotemporal trajectory data all bring resistance to the mining and analysis of co-occurrence patterns.Based on a variety of spatio-temporal trajectory data,this work aims to use advanced methods embracing data mining,natural language processing,visualization and so on,to deeply explore urban human mobility and attempt to answer three important questions: how to define and understand co-occurrence patterns,how to mine co-occurrence data effectively and efficiently,and how to analyze co-occurrence patterns.First of all,this paper understands cooccurrence patterns from the perspective of frequent patterns,and provides a quantitative definition of co-occurrence patterns.Then to mine co-occurrence data,this research proposes an effective co-occurrence pattern mining algorithm inspired by the basic idea of layer-by-layer iterative search and an efficient mining method based on FP-Tree.Finally,for the analysis of co-occurrence patterns,this paper analyzes the co-occurrence pattern from a global perspective and a regional function perspective,respectively,and presents a series of visual design to reduce the difficulty of analyzing co-occurrence patterns.Based on Shanghai taxi data,POI data,road network data,etc.,the performance of the proposed method is verified.The experimental results demonstarte that the proposed method is superior to the understanding and definition of co-occurrence pattern,the effective and efficient mining of co-occurrence data,and the analysis of co-occurrence patterns.