Unsupervised feature selection via fuzzy c-means clustering and binary atom search algorithm

This paper presents a proposed algorithm that combines the Binary Atomic Search (BAS) algorithm and the Fuzzy C-means Method (FCM) technique, called BAS-FCM. The proposed algorithm, BAS-FCM, is based on two-stage data processing. The first stage involves selecting important features from the datasets and discarding unimportant ones using the BAS algorithm, which relies on a proposed fitness function to test feature sets and select the best ones for classification and clustering. The second stage involves classifying an unsupervised datasets using FCM. This innovative combination leverages both features to improve accuracy, efficiency, and usefulness on high-dimensional datasets. Experiments on five different datasets demonstrate that the proposed method consistently achieves higher scores for silhouettes with fewer selected features compared to the default clustering technique. These results demonstrate the effectiveness and robustness of the proposed BAS-FCM method compared to the classical method, making it a promising approach for improving the performance of clustering-based feature selection.