Multi-strategy artificial protozoa optimizer for unconstrained functions, engineering problems, and feature selection

This paper presents a novel method for global optimization and feature selection utilizing an innovative version of the artificial protozoa optimizer, which was motivated by artificial protozoa. Feature selection improves classification by removing redundant and unnecessary characteristics and is an important field of research in artificial intelligence. This optimizer has proven to be easy to use, adaptable, and has limited coefficients in a variety of optimization problems. However, it has some flaws, including poor convergence competency and a tendency to get stuck at local extremes in various optimization problems. Consequently, this work presents an innovative artificial protozoa optimizer that increases the efficiency of the basic artificial protozoa optimizer through several innovative procedures: First, identifying individual solutions using Halton sequencing, thus increasing the diversity of the population. Second, the basic optimizer is modified to alter its search capabilities using an improved solution quality operator. Third, the proposed optimizer is prevented from getting stuck in local optimums by supplementing the optimal solution with a t-distribution perturbation. Finally, an elitist learning strategy was utilized to the best global individual to help jump out of local optimums when the search process is identified to be in a convergence case. The proposed algorithm was evaluated on the 10 complex test functions of the global CEC 2019 test suite and the 10 complex test functions of the global CEC 2020 test suite to examine its search performance and determine its effectiveness. The applicability of the proposed algorithm was used to extensively optimize six structuring design problems under various constraints. Furthermore, the efficiency level of the proposed algorithm in feature selection problems was evaluated using a collection of 24 publicly available datasets, and its results were contrasted with those of a group of well-known peers.