High-order conditional mutual information maximization for dealing with high-order dependencies in feature selection

This paper presents a novel feature selection method based on the conditional mutual information (CMI). The proposed High Order Conditional Mutual Information Maximization (HOCMIM) incorporates high order dependencies into procedure and has straightforward interpretation due to its bottom-up derivation. HOCMIM is derived from CMI's chain expansion expressed as maximization optimization problem. problem solved using greedy search procedure, which speeds up entire process. experiments are run set of benchmark datasets (20 in total). compared with eighteen state-of-the-art algorithms, results two supervised learning classifiers (Support Vector Machine K-Nearest Neighbor). achieves best terms accuracy shows be faster than counterparts.