Enhancing the Parkinson’s Disease Detection Through Machine Learning and Feature Engineering

Abstract

The main objective of this study is to generate a common approach which links machine learning models for prediction of Parkinson’s Disease (PD) having complications with the movement and other non-motor features. We considered PD diagnostic accuracy, as the early and precise identification play a vital role in management therefore, we explored different traditional machine learning algorithms; Logistic Regression, K-Nearest Neighbor, Support Vector Machine, Gaussian Naïve Bayes, Decision Tree, and Random Forest Classifiers in order to determine how they could be used for predicting the disease. Addressing data imbalance, the Balanced Random Forest was an approach implemented to guarantee a fair data class representation. Our approach consists of three major stages starting with data selection and cleaning then processing, model training, and validation that allows us to comprehensively uncover and utilize data-driven insights to improve the precision and predictive power of our models. This paper illustrates step by step method including the application of feature engineering, performance optimization, and the strategic implementation of ensembles to overcome the challenges of PD diagnosing. It is suggested that we could be able to really make progress in early detection of Parkinson’s Disease cases with the help of machine learning methods, which may represent novel research perspectives in the future and could be used in clinical practice.

Topics & Concepts