Voice feature-based diagnosis of Parkinson’s disease using nature inspired squirrel search algorithm with ensemble learning classifiers

• Published in: Iran Journal of Computer Science (Online) Vol. 8; no. 2; pp. 511 - 535
• Main Authors: Shibina, V., Thasleema, T. M.
• Format: Journal Article
• Language: English
• Published: Cham Springer Nature B.V 01.06.2025
• Subjects: Classification; Decision trees; Degeneration; Diagnosis; Ensemble learning; Kernel functions; Machine learning; Parkinson's disease; Principal components analysis; Radial basis function; Search algorithms; Squirrels; Support vector machines
• ISSN: 2520-8438; 2520-8446
• DOI: 10.1007/s42044-025-00232-0

Parkinson’s disease is a neurological condition primarily affecting individuals aged 55–65, leading to impairments in movement and speech due to the degeneration of specific brain regions. Recent advancements in machine learning have demonstrated the potential of using acoustic signals for the early detection of Parkinson’s illness. This study proposes a robust classification framework for PD diagnosis by leveraging machine learning techniques with advanced feature selection methods. The proposed approach employs the squirrel search algorithm (SSA) for feature selection, combined with principal component analysis (PCA) and ensemble learning, to enhance the efficiency of conventional machine learning algorithms. To ensure robustness and reliability, hyperparameter tuning and tenfold cross-validation were applied during the evaluation process. The proposed framework utilizes different classification models, including decision trees (DTs), support vector machines (SVM) with linear and radial basis function (RBF) kernels, and random forest (RF). In addition, ensemble classifiers such as gradient boost, AdaBoost, stacking, extra tree, light gradient boosting machine (LGBM), CatBoost, and voting classifiers were implemented and compared. The experiments were conducted using a data set from the University of California, Irvine (UCI) repository, comprising 64 healthy controls and 188 PD patients. Among the evaluated methods, the extra tree classifier combined with PCA and SSA achieved the highest performance, with an accuracy of 97.19%, precision of 97.32%, recall of 97.19%, F1-score of 97.17%, and an AUC score of 0.98. This work highlights the effectiveness of the proposed framework in accurately diagnosing Parkinson’s disease, offering a promising tool for early detection and intervention.