Prediction of Parkinson’s Disease Using Ens...

Prediction of Parkinson’s Disease Using Ensemble Learning

Author Name : Rathi G, Akhila A, Amritha P Menon, Dhanushia A

ABSTRACT The idea is to provide a methodology to predict Parkinson’s disease by using the biometric voice signals from the patients, and extracting the voice features from these signals and then feeding it to an ensemble learning model. Parkinson’s disease is a neurological disorder that affects the cells of the substantia nigra of the brain. The disease has no definitive lab or imaging tests to diagnose it and it has no cure yet. It has 5 stages and the symptoms worsen at each stage. The changes in voice of the patients are one of the symptoms that occur in the early stages. The UCI ML Parkinson’s disease speech dataset contains the 23 features, extracted from the voice signals of both healthy people and Parkinson’s patients using Multi-Dimensional Voice Program (MDVP). The ensemble learning algorithm, XGboost (eXtreme Gradient Boosting) is trained on this dataset and is validated. The model predicts if the person has Parkinson's disease or not based on the features. This method provides better performance when compared to the other methodologies used on the medical imaging test dataset and the methodologies using voice features. It can also diagnose the disease in the earlier stages which is helpful in the treatment of the patient.