Research On Mining Adjoint Pattern Of Spatial-Temporal Trajectory Data In Cloud Computing Environment

As the the rapid development of satellite positioning technology, wireless communication and track detection device, people can easily collect huge amount of spatial-temporal trajectory data. The position and the attributes of the moving objects are likely to change over time. In order to evaluate the cause, to predict the trend in the future, people not only need to know an object’s properties and position information, but also want to know the cause and effect of the event about the object. Spatial-temporal trajectory data can show attributes of the moving objects effectively. It contributes to research on human behavior patterns, transport and logistics, animal behavior and marketing and so on by analyzing the spatial-temporal trajectory data of different objects. As an important content of data mining, mining patterns of spatial-temporal trajectory data has attracted many researchers. As an important part of the pattern of spatial-temporal data, adjoint pattern of spatial-temporal trajectory is widely used in discovering the objects with same pattern or similar pattern and in researching the relationship of different objects. In this thesis, our main research content is about the algorithm of mining adjoint pattern of spatial-temporal trajectory data, and the main research achievement are as follows:(1) Propose an algorithm for mining adjoint pattern of spatial-temporal trajectory based on the grid index(MAP-G). With the help of grid index, we can not only improve the speed of searching the candidate objects, but also simplifying the trajectory data, reducing the amount of calculation to improve the efficiency of the algorithm. The experimental results demonstrate that the algorithm I proposed is more efficient in searching the candidate objects than the algorithms which searching candidate objects using DBSCAN, and the accuracy is higher as my algorithm can ignore some inaccurate results.(2) Present a parallel algorithm which called algorithm P-CMC, mining adjoint pattern of spatial-temporal trajectory data based on the algorithm CMC. Using Map/Reduce parallel programming model implement it. To improve the efficiency of the algorithm by processing different data set with algorithm DBSCAN using different compute nodes. The experimental results demonstrate that the algorithm I proposed is much more efficient compared to algorithm CMC. The most important thing is that, with the increase of number of compute nodes, the P-CMC algorithm shows high efficiency obviously. And the P-CMC algorithm shows increasingly obvious advantages with the rapidly expanding data.