基于移动通信数据的用户移动轨迹预测方法

USERS MOBILE TRACK PREDICTION METHOD BASED ON MOBILE COMMUNICATION DATA

随着无线移动通信设备的发展,获取用户位置的手段更加多样,如何对轨迹进行建模并预测人类行为成为研究热点。现有方法主要针对GPS轨迹等连续轨迹进行建模预测,而对移动通信场景中人行为轨迹预测方法尚未研究。针对移动话单数据这种离散程度极大的轨迹数据建模,提出Match算法对人类轨迹进行预测。实验证明,85%的人类轨迹可以利用该算法正确预测。在此基础上,提出轨迹合并的方法,进一步提高了预测的准确率,并发现人类在以天为单位的尺度上,有30%的行为是自相似的。

With the development of wireless mobile communication devices,there are diverse means to obtain users＇ location,and the ways to model the track as well as to predict the human behaviours become the focus of the research.Existing means are mainly aiming at continuous trajectory like GPS track to model and predict,but for predicting human behaviour tracks in the scene of mobile communication,it is till the blank yet.In this paper,aiming at modelling the mobile calling list data,which is a kind of track data with very large discrete degree,we propose Match algorithm to predict human tracks.Experiment proves that 85% human tracks can be correctly predicted with this algorithm.On this basis,we then propose a method of tracks merging,which further improves the accuracy of prediction.Moreover,it is found that 30% of the human behaviours are self-similar taking the day as the scale of human beings.