Multi-Cluster Feature Selection Based on Isometric Mapping

This letter presents an unsupervised feature selection method based on machine learning. Feature is important component of artificial intelligence, learning, which can effectively solve the curse dimensionality problem. Since most labeled data expensive to obtain, this paper focuses method. The distance metric traditional algorithms usually Euclidean distance, and it maybe unreasonable map high-dimensional into low-dimensional space by using distance. Inspired this, combines manifold learning improve multi-cluster algorithm. By geodesic we propose a isometric mapping (MCFS-I) algorithm perform adaptively for multiple clusters. Experimental results show that proposed consistently improves clustering performance compared existing competing methods.