Low-Redundant Unsupervised Feature Selection based on Data Structure Learning and Feature Orthogonalization

Mahsa Samareh Jahani, Farid Saberi-Movahed, Mahdi Eftekhari, Gholamreza Aghamollaei, Prayag Tiwari

Abstract: Highlights•The proposed feature selection method, LRDOR, selects low-redundant features.•LRDOR uses the orthogonal representation idea to obtain uncorrelated features.•LRDOR defines a subspace distance via orthogonal representation of features.•LRDOR captures local correlation among features and data structure manifold.