Clustering With Outlier Removal

Cluster analysis and outlier detection are two continuously rising topics in data mining area, which fact connect to each other deeply. structure is vulnerable outliers; inversely, outliers the points belonging none of any clusters. Unfortunately, most existing studies do not notice coupled relationship between these tasks handle them separately. In this article, we consider joint cluster problem, propose Clustering with Outlier Removal (COR) algorithm. Specifically, original space transformed into a binary via generating basic partitions. We employ Holoentropy measure compactness without involving several candidates. To provide neat efficient solution, an auxiliary matrix introduced so that COR completely efficiently solves challenging problem unified K-means— theoretical supports. Extensive experimental results on numerous sets various domains demonstrate effectiveness efficiency significantly over state-of-the-art methods terms validity detection. Some key factors including partition number generation strategy application abnormal flight trajectory further analyzed for practical use.