DMM+C:一个融合多源数据的位置预测方法  

DMM+C:a Multi-source Data Fusion Location Prediction Method

作　　者：卢菁[1] 安吉 刘丛[1] LU Jing;Anjiri S.Nandwa;LIU Cong(School of Optical-Electrical and Computer Engineering,University of Shanghai for Science and Technology,Shanghai 200093,China)

机构地区：[1]上海理工大学光电信息与计算机工程学院,上海200093

出　　处：《小型微型计算机系统》2022年第11期2278-2284,共7页Journal of Chinese Computer Systems

基　　金：上海理工大学自然基金培育项目(20ZRPY08)资助。

摘　　要：Markov模型在位置预测领域有着广泛的应用,然而,动态Markov模型阶数难以确定.本文设计了一个在对用户地理轨迹和语义轨迹进行挖掘的过程中动态确定阶数的方法DMM+C,根据用户当前的轨迹序列和历史轨迹模式,结合轨迹前缀树(TPtree)以及相邻簇中用户信息,动态确定合适的阶数进行用户位置预测.该方法基于滑动时间窗口的PrefixSpan(STW-P)算法实现,通过挖掘最近一系列块中的移动模式以及用户位置访问频率不断更新用户的轨迹模式,捕捉用户移动序列模式随时间的变化规律,利用用户评论的概率分布提高位置预测精度.真实数据集上的实验证明了本方法的有效性.Markov model is widely applied in the field of location prediction.However,it is hard to decide the right order of the Markov model.We propose Dynamic Markov Model location prediction method(DMM+C),which determines the right order by combining the Multi-order Markov model with the trajectory prefix tree(TPtree).The method is designed based on an improved sliding time window-based PrefixSpan(STW-P) algorithm which constantly update the user′s trajectory patterns by mining the moving patterns in the most recent series of blocks and capture the change of the user′s mobile sequence patterns over time.We also combine user comments analysis to improve prediction accuracy.Experimental results on Foursquare dataset show that our proposed method is effective compared to the existing location prediction methods.

关 键 词：马尔可夫模型 语义轨迹 位置预测 频繁模式挖掘 用户评论 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程]