Feature selection based on chaotic binary black hole algorithm for data classification

With the advance of generating high-dimensional data, feature selection is the most significant procedure to guarantee selecting the most discriminative subset of features and to improve the classification performance. As a result, a binary black hole optimization algorithm (CBBHA) has been developed by getting inspired from natural phenomena. In this paper, the most discriminating features are selected by a new chaotic binary black hole algorithm (CBBHA) where chaotic maps embedded with movement of stars in the BBHA. Ten chaotic maps are employed. Experiments on three chemical datasets show the proposed algorithm, CBBHA, has an advantage over the standard BBHA in terms of selecting relevant features with a high classification performance. Additionally the performance of CBBHA is compared with BBHA in term of the computational time efficiency which is revealing that CBBHA outperforms the BBHA.