This paper focuses on the detection of trend-based knowledge contained in sequential data. Some key concepts are redefined and an algorithm using the inertia test to dynamically partition a sequence to recognize trend partitions is also proposed. Experiment results prove the effectiveness of the algorithm, and exposing the limitation of traditional primitives. Some suggestions on parameter selection in practical application are given at the same time.
更多
查看译文
关键词
Trajectory Data Mining,Clustering Algorithms,Novelty Detection,Feature Extraction,Probabilistic Databases
原文链接
分享
引用
