Parkinson Disease (PD) is one of the most common neurodegenerative disorders. It is a chronic disease that reduces dopamine fluid secretion in the brain causes the disorder of both motor and non-motor features. This paper intends to provide a comparative study on the performance measure of various popular machine learning algorithms on the PD dataset obtained from the University of California at Irvine (UCI) machine learning repository. It is observed that biasness prevails in the performance of the classifier towards the majority class due to the imbalanced class distribution of the PD dataset. Hence two most popular preprocessing techniques were employed to balance the dataset one being Synthetic Minority Oversampling TEchnique (SMOTE) and NEAR MISS (NM) an opposite to SMOTE. A SMOTE samples the minority class up to the level of majority class and NM downsamples and brings the majority class down to minority class. All the features in the dataset do not contain useful information about the dataset and also irrelevant data leads to false classification. So, feature reduction is done using information gain ratio and thus obtained reduced dataset is then subjected to classification. For classification five popular classifiers such as Random Forest (RF), Support Vector Machine (SVM), Naive Bayes (NB), K-Nearest Neighbor (KNN) and Decision Trees (DT) were used to compare the performance with the balanced and imbalanced dataset. The evaluation of the classifier’s performance is recorded in terms of accuracy, precision, recall, and F-Measure. The results of the conducted experiments show that balancing the majority and minority classes improve precision and recall and there is an increase in accuracy as well as precision. When compared with other classifiers, RF with SMOTE preprocessing was found to be prominent with the information gain greater than 0.18.