A Hybrid Chaotic Particle Swarm Optimization with Differential Evolution for feature selection

The selection of feature subsets has been broadly utilized in data mining and machine learning tasks to produce a solution with a small number of features which improves the classifier's accuracy and it also aims to reduce the dataset dimensionality while still sustaining high classification performance. Particle swarm optimization (PSO), which is inspired by social behaviors of individuals in bird swarms, is a nature-inspired and global optimization algorithm. Particle Swarm Optimization (PSO) has been widely applied to feature selection because of its effectiveness and efficiency. The PSO method is easy to implement and has shown good performance for many real-world optimization tasks. However, since feature selection is a challenging task with a complex search space, PSO has problems with pre-mature convergence and easily gets trapped at local optimum solutions. Hence, the need to balance the search behaviour between exploitation and exploration. In our previous work, a novel chaotic dynamic weight particle swarm optimization (CHPSO) in which a chaotic map and dynamic weight was introduced to improve the search process of PSO for feature selection. Therefore, this paper improved on CHPSO by introducing a hybrid of chaotic particle swarm optimization and differential evolution known as CHPSODE. The search accuracy and performance of the proposed (CHPSODE) algorithms was evaluated on eight commonly used classical benchmark functions. The experimental results showed that the CHPSODE achieves good results in discovering a realistic solution for solving a feature selection problem by balancing the exploration and exploitation search process and as such has proven to be a reliable and efficient metaheuristics algorithm for feature selection.