Binary Black Widow Optimization Algorithm for Feature Selection Problems

• Published in: Learning and Intelligent Optimization Vol. 13621; pp. 93 - 107
• Main Authors: Al-Saedi, Ahmed, Mawlood-Yunis, Abdul-Rahman
• Format: Book Chapter
• Language: English
• Published: Switzerland Springer International Publishing AG 2023; Springer International Publishing
• Series: Lecture Notes in Computer Science
• Subjects: Black widow optimization algorithm; Classification; Data mining; Feature selection; Metaheuristic algorithms
• ISBN: 9783031248658; 3031248651
• ISSN: 0302-9743; 1611-3349
• DOI: 10.1007/978-3-031-24866-5_7

In this research work, we study the ability of a nature-inspired algorithm called the Black Widow Optimization (BWO) algorithm to solve feature selection (FS) problems. We use the BWO as a base algorithm and propose a new algorithm called the Binary Black Widow Optimization (BBWO) algorithm to solve FS problems. The evaluation method used in the algorithm is the wrapper method, designed to keep a degree of balance between two objectives: (i) minimize the number of selected features, (ii) maintain a high level of accuracy. We use the k-nearest-neighbor (KNN) machine learning algorithm in the learning stage to evaluate the accuracy of the solutions generated by the BBWO. This study has two main contributions: (a) applying the BBWO algorithm to solve FS problems efficiently, and (b) test results. The performance of the BBWO is tested on twenty-eight UCI benchmark datasets and the test results were compared with six well-known FS algorithms (namely, the BPSO, BMVO, BGWO, BMFO, BWOA, and BBAT algorithms). The test results show that the BBWO is as good as, or even better in some cases than the FS algorithms compared against. The obtained results can be used as new a benchmark and provide new insights about existing FS solutions.