Supervised techniques for Parkinson’s detection

Abstract

In approximately the past 25 years, the number of people suffering from Parkinson’s disease doubled. As of 2019, a survey by the World  Health Organization (WHO) showed that 8.5 million people across the globe were affected. Parkinson’s disease is a brain disorder that  causes uncontrolled, unintended movements such as shaking, difficulty in balancing, etc. A subset of Artificial Intelligence, Machine  Learning algorithms process large data sets, identify patterns, learn from them, and execute tasks autonomously. Owing to the amount  of data generated by each patient in the healthcare department, the amalgamation of the two fields - Machine Learning and Healthcare,  led to great advancement in research and development. In this paper, we identified the presence of Parkinson's disease based on the  report of a given individual. The aim was to create a holistic approach for identifying the presence of the disease and determining the  best-suited algorithm by implementing and comparing various algorithms. In order to achieve this, three machine learning algorithms –  SVM, XgBoost and Random Forest were employed. On comparing the results, XgBoost proved most efficient with an accuracy of 92.308%,  recall of 94.340%, precision of 92.593% and F1 score of 96.154%.