Optimizing Feature Selection with Binary Hippopotamus Algorithm and Random Forest: A Case Study on Chronic Kidney Disease Classification Zuraidah Fitriah, Ratno Bagus Edy Wibowo, Syaiful Anam
Brawijaya University
Abstract
This paper proposes an integrated framework combining the Binary Hippopotamus Optimization Algorithm (BHOA) and Random Forest (RF) for feature selection and classification in chronic kidney disease (CKD) prediction. Evaluated on the UCI CKD dataset, the proposed method examines four different S-shaped transfer functions to guide the binary search process. Results show that transfer function T_(S_2 ) achieved the highest final fitness score of 0.9421 +- 0.0024, effectively balancing accuracy with feature sparsity, while transfer function T_(S_4 ) attained the highest f_1-score of 0.9545 +- 0.0208, emphasizing classification performance. These findings highlight the critical influence of transfer function design in binary optimization and demonstrate the potential of the proposed approach to build accurate and interpretable predictive models for medical decision support.
Keywords: Binary Hippopotamus Algorithm- Chronic Kidney Disease Classification- Random Forest
