期刊
IEEE ACCESS
卷 6, 期 -, 页码 4895-4903出版社
IEEE-INST ELECTRICAL ELECTRONICS ENGINEERS INC
DOI: 10.1109/ACCESS.2017.2778690
关键词
Trajectory mining; sensor data; IoT
资金
- CAS Pioneer Hundred Talents Program
- MOE Key Laboratory of Machine Perception at Peking University [K-2017-02]
- NRF - Korean Government [015R1D1A1A01058171]
- [C05407260100474320]
Trajectory mining is an interesting data mining problem. Traditionally, it is either assumed that the time-ordered location data recorded as trajectories are either deterministic or that the uncertainty, e.g., due to equipment or technological limitations, is removed by incorporating some pre-processing routines. Thus, the trajectories are processed as deterministic paths of mobile object location data. However, it is important to understand that the transformation from uncertain to deterministic trajectory data may result in the loss of information about the level of confidence in the recorded events. Probabilistic databases offer ways to model uncertainties using possible world semantics. In this paper, we consider uncertain sensor data and transform this to probabilistic trajectory data using pre-processing routines. Next, we model this data as tuple level uncertain data and propose dynamic programming-based algorithms to mine interesting trajectories. A comprehensive empirical study is performed to evaluate the effectiveness of the approach. The results show that the trajectories could be modeled and worked as probabilistic data and that the results could be computed efficiently using dynamic programming.
作者
我是这篇论文的作者
点击您的名字以认领此论文并将其添加到您的个人资料中。
推荐
暂无数据