Multi-cluster nonlinear unsupervised feature selection via joint manifold learning and generalized Lasso

Yadi Wang, Mengyao Huang, Liming Zhou, Hangjun Che, Bingbing Jiang

Abstract: Highlights•A novel method named MUNFS is proposed to remove irrelevant and redundant features.•MUNFS solves the imbalance between local and global structure retention of data.•MUNFS solves the issue of inadequate treatment of data nonlinear relationships.•The importance of each feature can be evaluated more accurately.•The superior performance of the MUNFS method is verified by numerous experiments.