Parkinson’s disease Detection Using Tree Based Machine Learning Algorithms

Parkinson’s Disease (PD), also known as primary Parkinsonism is a persistent, idiopathic, degenerative nervous disorder which results from lack of dopaminergic neurons in the substantia nigra pars compacta, which is the source of nigrostriatal dopamine pathway within the midbrain. The clinical detection relies on motor symptoms recognition. Significant neurological damage is already done by the time motor symptom occur. Early detection is necessary to stalk the progression of the disease. The problem of detection of PD comes under classification. Several tree-based classification algorithms were applied to the dataset retrieved from UCI machine learning database. The dataset was first split into train and test data. Various models were created using four different algorithms. Correlation coefficients were calculated for each of the features in the dataset. The model was fitted with train data obtained after removing highly correlated features. Predictions were made and various parameters were considered for comparison. Accuracy, precision, recall, F1-Score, Youden Index, error rate and specificity were the parameters calculated. Out of the four algorithms (Decision Tree, Random Forest, XGBoost and LightGBM), LightGBM achieved the highest accuracy of 97.43%.