Backward automatic screening feature selection algorithm based on multi-evaluation criterion fusion

In view of the tedious problem that the feature selection algorithm based on multi-evaluation criterion fusion needs to manually set the feature threshold repeatedly to obtain the optimal feature subset, a backward automatic selection feature selection algorithm based on multi-evaluation criteria fusion is proposed in this paper. In this algorithm, the feature evaluation value is calculated using multi-evaluation criteria fusion algorithm, then the discrimination performance of Support Vector Data Description (SVDD) is used as the screening criterion, and finally the optimal feature subset is obtained by backward automatic screening algorithm. The experimental results show that the feature selection algorithm has the advantages of automatic screening, noise robust, clutter robust, low feature dimension, high discrimination rate and low deterioration of discriminative performance for small sample problems.