In this project, several machine learning techniques was implemented on the dataset which contained speech data about parkinson disease. Some distinct methods were used for preprocessing and feature selection, such as LDA, ICA, PCA w/o whitening, Sequential Backward Feature Elimination and Autoencoders. For classification, two main categories of classifiers were used in this project, generative and discriminative. From generative classifiers, Optimal bayes classifier was used in addition to Parzen window or K-Nearest Neighbors for density estimation. In addition, Gaussian Mixture Model was used as alternative density estimation method. SVM, KNN, Logistic Regression, Decision Tree, and MLP were used as instances of discriminative classifiers. Furthermore, bagging and ensemble methods were used as an extra part. Results compared based on different criteria such as accuracy, f1 score, AUC, ROC Curve, and confusion matrix.