An Integrated XGBRF Framework for Early Identification of Parkinson’s Disease

Parkinson's Disease (PD) is a neurodegenerative disorder that affects movement, balance and coordination. It is a chronic disease that results from the degeneration of dopamine producing neurons in the brain. Several artificial intelligence and machine learning techniques has been introduced for the early detection of PD. Early detection of PD is indispensable for slowing down its progress and providing the possibility to access the therapy. Therefore, the research focus on developing a hybrid framework (XGBRF) utilizing XGBoost and random forests to enhance the accuracy of predicting the PD is proposed. It is hypothesized that the proposed framework will benefit from this complementary characteristic by using RF's strength in variance reduction through bagging and XGBoost on bias minimization through boosting. This research was tested on publicly available datasets using cross-validation with hyper-parameter tuning for the optimization of optimal performance. In comparison, the proposed framework results in higher accuracy rate of 0.97, precision of 0.92, F1 score of 0.96 and AUC-ROC value of 0.93 than the existing classifiers. Therefore, the outcome of this research would essentially mean that the proposed framework may constitute a solid framework for medical diagnostics and serve as a better prognosis of PD.