A Hybrid Ant Colony Optimization for Parkinson’s Disease Classification Based on Synthetic Minority Oversampling and Adaptive Synthetic Techniques

Parkinson’s disease (PD) is a neurological ailment of the nervous system, first characterized in 1817. Around 90% of PD patients experience speech problems. Therefore, early detection and identification of PD are critical for disease progression and treatment. This study proposes a unique approach for detecting PD using features extracted from speech signals of patients with PD. We categorize the hybrid machine learning (ML) technique into two stages: (i) pre-processing of data, followed by (ii) hybrid ant colony optimization (ACO), and (iii) classification using the extra tree, cat boost, and Xgboost. However, the PD data set’s class distribution is skewed (unbalanced). Therefore, we applied the Synthetic Minority Oversampling Technique (SMOTE) and Adaptive Synthetic technique (ADASYN) to balance this unbalanced data. We classified the PD dataset using the hybrid ACO with extra tree, hybrid ACO with cat boost, and hybrid ACO with Xgboost classification methods after transforming it to a balanced majority class. The suggested hybrid methods, which combine SMOTE-ACO+extra tree, and ADASYN-ACO+Extra tree gave classification accuracy rates of 99.28%, and 95.72%. The results indicated that the proposed hybrid ACO technique produced promising results in PD classification.