Parkinson's disease is an incurable aural derangement caused by nerve cell destruction in the human brain. It is essential to identify the Parkinson's disease so that medical intervention can be made to prevent the patient condition from worsening. There are several attributes present in the dataset which is used to predict Parkinson's disease using Data Mining Algorithms. Dataset with many features may increase the complexity of analysis and it may degrade the Model comparison. This research suggests a best feature selection technique that chooses important features which have a favorable effect on the model's comparison. Frequently two types of attributes Selection techniques used in machine learning, Namely Filter-based and Wrapper-based feature selection. This paper compares the comparison of several classifiers using the best features obtained by various feature Selection techniques. Classifiers such as Decision Tree, SVM, KNN, Gradient Boosting and feature Selection algorithms such as Mutual information gain (Under Filter based feature Selection technique), Random mutation hill climbing, and Mutual information based cuckoo search optimization (Under Wrapper based feature Selection technique) has been used for comparison. The results of the experiment show that the application of feature Selection can improve the comparison of the model.

:Filter based, Wrapper based, feature Selection, Mutual information gain, Random mutation hill climbing, Cuckoo search optimization, KNN, SVM, Gradient Boosting, Decision tree, classifiers.