Diagnosis of Parkinson’s disease based on voice signals using SHAP and hard voting ensemble method

• DOI: 10.1080/10255842.2023.2263125
• Publication Date: 2023-09-29T06:35:07Z
• Authors: Paria Ghaheri; Hamid Nasiri; Ahmadreza Shateri; Arman Homafar
• Keywords: Male; Signal Processing (eess.SP); FOS: Computer and information sciences; Computer Science - Machine Learning; Parkinson Disease; Signal Processing, Computer-Assisted; 004; 3. Good health; Machine Learning (cs.LG); Machine Learning; 03 medical and health sciences; 0302 clinical medicine; Audio and Speech Processing (eess.AS); Voice; FOS: Electrical engineering, electronic engineering, information engineering; Humans; Diagnosis, Computer-Assisted; Voting; Electrical Engineering and Systems Science - Signal Processing; Algorithms; Electrical Engineering and Systems Science - Audio and Speech Processing

Parkinson's disease (PD) is the second most common progressive neurological condition after Alzheimer's. The significant number of individuals afflicted with this illness makes it essential to develop a method diagnose conditions in their early phases. PD typically identified from motor symptoms or via other Neuroimaging techniques. Expensive, time-consuming, and unavailable general public, these methods are not very accurate. Another issue be addressed black-box nature machine learning that needs interpretation. These issues encourage us novel technique using Shapley additive explanations (SHAP) Hard Voting Ensemble Method based on voice signals more accurately. purpose study interpret output model determine important features diagnosing PD. present article uses Pearson Correlation Coefficients understand relationship between input output. Input high correlation selected then classified by Extreme Gradient Boosting, Light Boosting Machine, Bagging. Moreover, weights determined performance mentioned classifiers. At final stage, SHAP diagnosis. effectiveness proposed validated 'Parkinson Dataset Replicated Acoustic Features' UCI repository. It has achieved an accuracy 85.42%. findings demonstrate outperformed state-of-the-art approaches can assist physicians cases.