Milling Significant Blowout Patterns
One blowout pattern is a zone on multiple data streams in which the data points are dense but highly unbalanced. It can be applied into a variety of fields. For example, it may imply a successful sales promotion especially for a few areas. The difficulties of exploring blowout patterns are: 1). It is not periodic;2). The distribution is unknown;and 3). How to distinguish it from outliers or clusters. We have proposed a novel method for it. First, we employ a density based clustering algorithm to detect dense data points. Second, we use a novel concept (Normal Standard Deviation: NSD) for the evaluation of the data distribution in the zones. The zones that include dense but unbalanced data points are highlighted as significant blowout patterns. The results demonstrate that our method can find and locate blowout patterns efficiently and effectively compared to kernel framework and slide window.
Blowout pattern multiple data stream data distribution normal standard deviation
Feng Chen
College of Information Science and Engineering Henan University of Technology Zhengzhou, China
国际会议
杭州
英文
177-180
2012-10-28（万方平台首次上网日期，不代表论文的发表时间）