Feature Selection and Classification Using CatBoost Method for Improving the Performance of Predicting Parkinson’s Disease

Abstract

Several studies investigated the diagnosis of Parkinson’s disease (PD), which utilized machine learning methods such as support vector machine, neural network, Naïve Bayes and K-nearest neighbor. In addition, different ensemble methods were used such as bagging, random forest and boosting. On the other hand, different feature ranking methods have been used to reduce the data dimensionality by selecting the most important features. In this paper, the ensemble methods, random forest, XGBoost and CatBoost were used to find the most important features for predicting PD. The effect of these features with different thresholds was investigated in order to obtain the best performance for predicting PD. The results showed that CatBoost method obtained the best results.