Binary Arithmetic Optimization Algorithm for Feature Selection in Pattern Recognition

Feature selection provides a technical way for pattern recognition to identify the important features from a dataset. However, Arithmetic Optimization Algorithm (AOA), which shows competence in solving continuous optimization problems, still cannot be employed in feature selection directly. In this paper, we propose a binary version of Arithmetic Optimization Algorithm (BAOA) that introduces a novel V-shaped time-varying transfer function to implement the transition from continuous search space to binary space. The proposed BAOA is proved to have superior performance in terms of convergence speed and ability to escape local optima. A set of evaluation indicators are employed to evaluate and compared the different algorithm over 15 datasets from the UCI repository. The simulation results prove the capability of the proposed binary version of the Arithmetic Optimization Algorithm to search for the optimal subset of features.