부산대학교 도서관

The feature selection process aims to obtain the vital information contained in the dataset. Determining the high-impact features has a key role in improving the classification process, applied in many scientific and medical fields within our study. This paper proposes a hybrid SD-BASO algorithm between the statistical dependence (SD) technique and the binary atom search optimization (BASO) algorithm. This algorithm depends on a proposed fitness function through which the essential features that affect the classification process are obtained. The experimental results on the datasets showed that the proposed algorithm, which we refer to as SD-BASO, is superior to the classical algorithm in terms of accuracy in the results and the number of features selected.