Extraction and Trend Detection of Unusual Patterns over Data Streams

In this paper, emphasis is put on a hot topic abnormal behavior analysis, named unusual pattern extraction and trend detection Contributions of this paper include: （1） A new framework ratio of feature strength for unusual pattern extraction is proposed, which provides a standard for unusual pattern mining; （2） Based on the extraction of unusual pattern, a method for trend detection of unusual pattern is also proposed by using least square regression method The performance study shows that the framework of unusual pattern measurement and algorithm is reasonable, and the proposed trend detection method is executable and efficient