中文摘要：

针对海量的用户轨迹数据进行研究，提出一种动态分析移动对象轨迹模式、预测轨迹位置的方法（PRED）。首先使用改进的模式挖掘模型，提取轨迹频繁模式（简称T一模式）；然后提出DPTUpdate算法，设计蕴涵时空信息的快捷数据结构——DPT（dynamic pattern tree），存储和查询移动物体的T-模式，并提出Prediction算法计算最佳匹配度，得到移动对象轨迹的预测位置。基于真实数据集进行对比实验，结果证明，PRED方法可提供动态分析的能力，平均准确率达到72％、平均覆盖率达到92．1％，与已有方法相比，其预测效果有显著提升。

英文摘要：

According to the research about a large amount of users＇ trajectory data, this paper proposed a method （PRED） to analysis trajectory pattern dynamically and predict location. The first step was using the improved pattern mining model to extract trajectory frequent patterns （named T-pattern）. Then it put forward the DPTUpdate algorithm to design a data structure： DPT, which contained spatio-temporal information and could store and query the trajectory frequency pattern of moving objects. In addition, it presented the Prediction algorithm to calculate the optimal matching degree and got the predicted location of moving object trajectories. According to a series of contrast experiment based on the actual data, the PRED method can provide real-time analysis during the process, the average accuracy of this method reaches 72% and the average coverage reaches 92.1%. The prediction effect has been increased significantly comparing with existing methods.