Binary hippopotamus algorithm with random forest for optimizing feature selection problem

This research introduces Binary Hippopotamus Optimization Algorithm (BHOA) that operates in binary space to solve binary optimization problems. BHOA is inspired by the behavior of hippopotamuses in the wild. BHOA implements the diverse roles within hippopotamus herds and exhibits strong exploration and exploitation capabilities. BHOA uses the Levy movement which makes BHOA effective in solving feature selection problems. Feature selection is crucial in optimization, as irrelevant features increase computational time and reduce performance. Random Forest (RF) is machine learning model known for its speed and efficiency. RF serves as an excellent tool for evaluating the selected features. The results show that BHOA consistently outperformed other algorithms across all tested datasets, achieving the highest $ f_1 $-score with quicker convergence. On the dry bean dataset, it reached 0.9237 by the 28th iteration, maintaining stability until the 30th. For the red wine quality dataset, BHOA achieved the highest $ f_1 $-score at the 29th iteration, outperforming others like CBBA and BBA. On the breast cancer, it demonstrated efficiency with 0.9824 by the 20th iteration, while other algorithms lagged. Similarly, for the mobile price classification, BHOA achieved 0.9441 at the 29th iteration, with competitors showing slower convergence and lower performance. These results confirm BHOA's superiority in binary feature selection across diverse datasets.