Parkinson’s Disease Detection and Classification Through Gait Analysis

Parkinson’s disease (PD) is a novel neurodegenerative condition that exhibits both motor and non-motor symptoms. Gait problems are a common occurrence for PD patients in the early stages of the illness. So, analysis structures primarily based on gait issues are at the leading edge on latest PD detection research. Machine learning frameworks can be proposed based totally on various classifier models to categorize Parkinson’s disease (PD) the usage of units of gait capabilities. There are frameworks which might be employed for the aggregate of diverse function units, and they have difference in phrases of combining feature units. Our focus in this research paper is to train various machine learning models for Parkinson’s disease classification and evaluate the appropriateness of the models for the chosen dataset. We have performed univariate and multivariate analysis for exploring the inherent characteristics of the dataset. We have fine-tuned and optimized three machine learning models to optimally fit to the undertaken Parkinson’s dataset—random forest, gradient boost, and XGBoost. It has been observed that XGBoost performed well when evaluated against the metrics “accuracy, sensitivity, specificity, and precision”. Gradient boost algorithm demonstrated better performance when evaluated against the F1-score metric. Overall, XGBoost model outperformed the remaining two learning models for the Parkinson’s dataset with an accuracy of 98.1%.