Detection of Parkinson's Disease using Extreme Gradient Boosting

Abstract

Parkinson's disease is a brain-related disease that is common in every person mainly persons above age 45 years. This disease causes numbness in muscles, swallowing problems, bending of the back, shivering in hands, smell dysfunction, speaking problem, Hearing problem, and many more. Parkinson's disease has to be diagnosed as early as possible since the clinical tests, which take hours to detect, may cost a loss of time and money. An automated model for detecting Parkinson's disease in a person with greater accuracy is proposed in this paper. While several models for detecting Parkinson's disease have been established, they are all less reliable and precise. Our model is created using the gradient boosted decision tree, which not only reliably predicts Parkinson's disease in a human, but also predicts it quickly. The feature set contains 22 parameters of the voice signal, which are given to the XGBoost classifier. The developed model predicts Parkinson's disease with 96.6% of accuracy, 95.6% of sensitivity, 100% of specificity, 100% of Precision, F-Score 97.7%.