Data-Driven Prediction of Parkinson’s Disease Using Random Forest for Improved Clinical Decision Support Systems
Year: 2026
Type: conference-paper
Abstract: Parkinson’s Disease (PD) is a neurological disorder that is chronic and progressive in nature, mainly affecting motor control and speech-related functions. Treatment and management of this condition involves accurate diagnosis and in good time. This study uses a machine-learning model that uses the Random Forest algorithm in order to predict the existence of Parkinson’s Disease using the voice-based biomarkers. Genetic Algorithm is used to improve the performance of the models by feature selection and Bayesian Optimisation is used to optimise the hyperparameters. The accuracy of the proposed framework is 94% with ROC-AUC of 0.97, which is a good predictor. Compared to the current models, it can be seen that although the standard models like Random Forest and Support Vector Machines often claim accuracy ranging between 93.84% and 96.61, the use of Genetic Algorithms based feature selection and Bayesian Optimisation led to the model being much more efficient and robust. Such results suggest that AI-based diagnostic methods that are not invasive could be utilized to diagnose Parkinson’s Disease at its initial stages. The next research can be focused on integrating deep learning models, longitudinal voice data, and incorporating other clinical features to enhance the generalisation of the model and enable the detection at the earliest stage.
Language: English
Subfield: Health Information Management
Field: Health Professions
Domain: Health Sciences
Open Access status: hybrid
OpenAlex ID: W7203695802