MPCASMA: A Multi-Population Chaotic-based Hybrid Algorithm for Global Optimization and Its Application in Feature Selection

Contemporary, a significant volume of data is being produced, which has caused various problems comprising an exponential increase in data processing time, decreased accuracy of existing techniques, and increased demand for required hardware resources for data processing tools and methods. Feature selection deals with the problems by selecting the optimal subset of related features. Scholars proposed various methods for feature selection, the most recent of which is the use of meta-heuristic algorithms. Considering that new and voluminous datasets are being developed, this field is still one of the hot research fields. In this paper, an innovative optimization algorithm called MPCASMA is presented. In MPCASMA, Arithmetic Optimization Algorithm (AOA) and Slime Mold Algorithm (SMA) are hybridized through a multi-population strategy. The Sine, Leibovitch and Circle chaotic maps are adopted in the following, and a novel hybrid chaotic map is provided. Additionally, an advanced neighborhood search strategy is presented. Then, six different S-shape, U-shape, and V-shape transition functions are employed to obtain the most optimal binary variant of MPCASMA. Ultimately, twenty-three unimodal, multimodal, and fixed dimensions standard benchmark functions and eight real-world datasets are utilized to check the efficiency of the contributions and the superiority of the proposed method. The proposed algorithm is compared with the AEO, AOA, DO, EO, MRFO, SMA and TSA algorithms numerically and visually in the experiments. The results indicate the proposed algorithm's effectiveness and superiority over competitors.

Keywords:

Arithmetic Optimization Algorithm , Slime Mold Algorithm , Numerical Optimization , Feature Selection , K-Nearest Neighbor