Comparative Analysis of Artificial Rabbit Optimization and Ox Optimization Based Feature Selection for Parkinson’s Disease Diagnosis

Parkinson’s disease (PD) is a neurodegenerative disease that causes the deterioration of human motor functions, making early and accurate diagnosis important. Nowadays, machine learning is effective in diagnosing this disease using patients’ voice and image data. However, in high-dimensional biomedical data such as PD, the performance of machine learning algorithms largely depends on the quality and suitability of the input features. The aim is to determine the most informative subset of features that will enhance input quality through feature selection. In recent years, meta-heuristic algorithms have gained superiority in feature selection. In this study, a comparative analysis of two new nature-inspired metaheuristic algorithms, Artificial Rabbit Optimization (ARO) and Ox Optimization (OXO), is presented with the aim of selecting the most suitable feature subsets for PH diagnosis. Adaptations of these algorithms for feature selection have been made, and some modifications have been applied. The performance of ARO and OXO has been tested on the well-known Max dataset derived from sound measurements. The effectiveness of the feature subsets selected by ARO and OXO was evaluated using SVC, KNN, RF, and XGBoost. The results showed that both algorithms were able to significantly reduce the original set of 22 features (to 7-13 features) and markedly improve classification accuracy (up to 97.44%) compared to the baseline, indicating that they found effective feature subsets for PH diagnosis.