A Study On Data Mining Of Moving Object Trajectories

Trajectory data produced by mobile devices record spatial and temporal information on moving objects. The rapid development and vast application of positioning, wireless communication and mobile internet technologies have enabled us to easily obtain trajectories of moving objects, which has resulted in huge amount of trajectory data.Trajectory data contain rich information, whereby people can not only analyze characteristics of trajectories such as speed and direction of movement, but can also investigate behavior of moving objects such as the location where they always arrive or stay, as well as surrounding environment properties such as traffic and points of interest. Therefore, the potential of applying trajectory data is great. Currently, the data have been widely applied to various sectors such as transportation, public security, tourism, logistics and meteorology.Trajectory data, characterized by large amount of volume, variety of types and fast growth of size, are important data resources. With arrival of the big data era, there is an urgent need to apply new methods to trajectory data mining theory and technology so as to benefit the sustainable social and economic development.Based on trajectory data collected by 13,700 taxies from Shanghai Qiangsheng Holding Company in April 2015 and other data sources such as road network dataset, this dissertation conducted the research in three aspects as follows:(1) Design and development of cloud-computing-based trajectory data analysis platformThis dissertation designs and develops a trajectory data analysis platform based on cloud computing. By employing Hadoop HDFS as the distributed file management system, Spark as the distributed computing framework, and secondary development interface provided by the distributed computing system, the platform develops application software with various functions including exploratory analysis and semantic enhancement of trajectories, analyses on taxi operation, travel behavior of residents and road traffic characteristics. The results of tests show that the distributed computing system is superior to the stand-alone computing system in terms of computing time. With increasing data volume, the former is found far more efficient than the latter. For instance, when extracting taxi passenger information with 0.11 billion points, it took the distributed computing system (6 nodes) 6 minutes to finish the task, whereas a single computer spent 12.4 hours, which was 124 times of the distributed system.(2) Research on processing and analyzing trajectory dataThis research concentrates on development of methods in map-matching trajectory data and analyzing stops of trajectories (that is, important places where the moving object has stayed for a minimal amount of time).This study proposes a sequence-based map-matching method that simultaneously matches all points in a sequence. First, road segments are chosen as candidates for map-matching according to the distance threshold. Next, following network connectivity, topologically connected road segments are obtained. The shortest path algorithm is then used for determining road segments. This can significantly improve the accuracy of the matching task.Focusing on characteristics of trajectory data, this dissertation develops a method for detecting stops based on two types of velocity. First, consecutive points are examined to determine whether they are clustering points according to the instant velocity threshold. Next, stops are identified based on average velocity, the shortest time interval and instant velocity thresholds, among which the instant velocity threshold should be greater than the average velocity threshold. This method can partly solve the false negative problem resulting from outliers with extremely high velocity.(3) Application of taxi trajectory dataThis research focuses on analyses of taxi operation, travel behavior of residents and road traffic characteristics in neighborhoods.Based on taxi trajectory data during the month of April in 2015, characteristics of taxi operation, including average daily operation time, average daily operation mileage, average daily passenger capacity, average daily passenger time, average daily passenger mileage, total passenger time and passenger mileage, as well as empty-run time ratio and empty-run mileage ratio.Based on passenger trajectory data form April 20 to 26 (Monday to Sunday) in 2015, travel characteristics of residents are analyzed, including travel time, location, direction and association between getting on and off taxies.Based on trajectory data of a typical working day (Monday) and Sunday in April 2015, driving velocity is calculated for main roads, secondary roads and tertiary roads within outer ring of Shanghai, whereby temporal and spatial variation of the velocity are further analyzed by type of road, time of day and day of week.