Parkinson’s disease (PD) is a brain disease caused by death or damage of dopamine producer brain cells. In such case, the brain can not perform its normal functions. PD negatively affects human movements such as speech, walking and writing. In the diagnosis of this disease, detailed medical history, history of treatment, physical tests and some blood tests and brain films are required. Because these operations can be costly and difficult, less costly and easier making of the diagnosis has such important in this subject. In this study, it was aimed to diagnose PD with voice data from 252 individuals to support the doctor’s decision. In order to get better results, some pre-treatments were applied. The datas were balanced and datas that taken to treatments with systematic sampling method were determined. With the feature selection algorithm, some data groups were created by calculating the effect of the attributes on the label. Of the classification algorithms; Decision tree, Support Vector Machines and K Nearest Neighbor Algorithm are used and Performance evaluation criteria such as Accuracy, Sensitivity, Specificity, F-Measurement, Kappa, Auc - were evaluated. The data group with the highest performance value and the used classification algorithm were determined and model was created by using support vector machines algorithm and from 45% of data set that was sorten from the most effective to the most ineffective. The performance criterias has an accuracy of 85% besides in other criterias possitive results were earned. Thus, it was understood that medical decision support to the doctor would be provided with the help of applied model and data set formed by sound recordings of the individuals who possibly been PD