Diagnosis of Parkinson's Disease using Hybrid Ensemble Technique
Raj Sinha, Navpreet Kaur, Sandeep K. Gupta, Padmanabh Thakur
Abstract
Parkinson's disease (PD) presents a substantial global health challenge characterized by difficulties in movement, coordination, and muscle regulation. Diagnosis primarily relies on identifying symptoms like shaking, stiffness, slow movements, and trouble walking. PD can also impact speech, leading to changes in breathing, vocal quality, and volume. Research suggests that alterations in speech could be an early indication of PD, motivating the use of a specialized Parkinson's dataset for speech analysis in this specific research. The research investigates the use of diverse machine learning models, such as Random Forest classification, K Nearest Neighbors, XGBoost, and Support Vector Machine, combined with Principal Component Analysis (PCA) to assess and diagnose PD based on speech characteristics. These models demonstrate promising accuracy levels ranging from 80% to 91.5% when tested on the PD dataset. Ongoing research endeavors aim to develop more efficient combined models to improve overall performance. By amalgamating these models, the study achieves a peak accuracy of 91.5%, enabling precise identification of the disease.