Detection of Parkinson's Disease Using Machine Learning: A Case Study

Detecting Parkinson's disease is a complex task that comprises of examining different aspects of the body, mind, and the nervous system. But researchers have discovered a new way to use speech metadata to diagnose people with Parkinson's disease by extracting features from people's voices. In this paper, we have applied Naive Bayes, Logistic Regression, Gradient Boosting, Random Forest, Multilayer Perceptron and Stacked Classifiers to evaluate different learning models for detection and the Stacked Classifier and Gradient Boosted models reached the highest accuracy with 89.47% and 88.15%, respectively. The multilayer perceptron model has the lowest accuracy with 84.21%. Therefore, the Stacking Classifier and Gradient Boosting models were found to be optimal for similar data. This study provides insight into the selection of an appropriate machine learning models for accurate prediction.