Fuzzy multi-neighborhood entropy-based interactive feature selection for unsupervised outlier detection

Siyu Yang, Zhong Yuan, Chuan Luo, Hongmei Chen, Dezhong Peng

Abstract: Highlights•This paper considers the distribution characteristics of data and builds fuzzy multi-neighborhood granule to handle heterogeneous unbalanced data.•Fuzzy multi-neighborhood entropy and its related concepts are defined for uncertainty measures.•An unsupervised feature selection algorithm is constructed by ranking the importance of features according to relevance, redundancy, and interactivity.•The proposed algorithm is applied to outlier detection of unbalanced data and the comparison results show that it has great performance.