In this work, the accuracy of two machine learning algorithms including the SVM and Bayesian networks were investigated as two important algorithms in the diagnosis of the Parkinson’s disease (PD). We used the PD data in the University of California, Irvine (UCI). In order to optimize the SVM algorithm, different kernel functions and C parameters were used, and the results obtained showed that SVM with C parameter (C-SVM) with an average accuracy of 99.18% with the polynomial kernel function in the testing step had a better performance compared to the other kernel functions such as RBF and sigmoid as well as the Bayesian network algorithm. It was also shown that the ten important factors involved in the SVM algorithm were Jitter (Abs), Subject #, RPDE, PPE, Age, Shimmer APQ 11, NHR, Total-UPDRS, Shimmer (dB), and Shimmer respectively. We also proved that the accuracy of our proposed C-SVM and RBF approaches was in direct proportion to the value of the C parameter such that with increase in the amount of C, the accuracy in both kernel functions increased. However, unlike polynomial and RBF, sigmoid had an inverse relation with the amount of C. Indeed, by using these methods, we can find the most effective factors common in both genders (male and female). To the best of our knowledge, there has been no study on PD for identifying the most effective factors common in both genders.

Keywords