Anomaly Urban Mobility Detection And Visualization Design Based On Large Trajectory Data

With the rapid development of mobile devices,the scale of trajectory data is growing.Urban mobility study has gradually become a hot issue in recent years.It has important value in urban planning,land use,urban traffic management and other research fields.Urban mobility study mainly includes two parts: functional region identification and mobility pattern discovery.The method of functional region division and semantic functions identification are important components of functional region identification.At present,the traditional methods of dividing functional areas do not consider the spatial distribution of POI in cities,thus,the traditional methods of dividing functional areas have limitations.Anomaly mobility pattern detection is one of the most important problems in mobility pattern discovering.The existing methods for detecting anomaly mobility patterns can only determine whether a pattern is anomalous or not.However,according to different travel activities of people,there exists several different patterns of mobility distribution in cities.Therefore,the existing anomaly detection methods have limitations.This thesis aims at functional region identification and anomaly mobility pattern detection.In order to overcome the limitations of traditional methods,two different methods of functional area division are proposed in this thesis.They are called "density clustering method" and "Tyson polygon method".Density clustering method generate POI clusters by using density-based clustering algorithm.And the geometric shape of functional regions is generated by extracting the range of POI in each cluster.Tyson polygon method is based on the Tyson polygon generated by bus stations in cities and then generates urban areas.Based on POI data sets,a weighted TF-IDF method is constructed to identify the semantic functions of the generated functional areas.When detecting the anomaly mobility patterns,this thesis constructs the data set of urban movement distribution vectors according to the unit time period in different time scales,and creatively applies the clustering algorithm to generate clusters by these multi-dimensional vectors.Each cluster corresponds to a mobility pattern.Finally,the frequency of each pattern is used to determine whether it belongs to the anomaly mobility pattern or not.Areal trajectory data set and POI data set from the Wuhan City are utilized to verify the proposed method in this thesis.The experiment shows that the proposed method identifies the functional regions quickly and effectively identify,it can not only detectserval different anomaly mobility patterns,but also capture all the time intervals of each anomaly mobility pattern.By comparing two different methods in functional area division,Tyson polygon method can generate more accurate results.In addition,this thesis designs and implements a real-time interactive visualization system,which is used to show the identified functional regions and mobility distribution to users in a more intuitive way.