A Class Imbalance Learning Approach to Build an Efficient Machine Learning Model for the Diagnosis of Parkinson's Disease

Parkinson's disease is a progressive neuro-degenerative disorder that affects the central nervous system. Symptoms progress gradually and clinical diagnosis is normally possible only after symptoms have significantly progressed. Machine Learning(ML) models, which treats the diagnosis process as a binary classification problem, have been widely employed to diagnose PD. While applying ML algorithms to diagnose disease with low prevalence rate such as PD, one of the challenges that affects the accuracy of prediction is that most datasets will be imbalanced due to the practical difficulties in getting data from the patients. As a result, both classes will typically have unequal numbers of data points. In order to increase the performance of ML models dealing with an imbalanced dataset to detect PD, this research suggests incorporating the class imbalance learning technique. With normal training and class imbalance learning, five machine learning models-decision tree, K-nearest neighbour, Naive Bayes, support vector machine, and random forest-have each been individually assessed. While applying class imbalance learning, a considerable improvement in performance has been seen in each model. This technique may be used to accurately detect other similar diseases with low incidence rates.