The classifier performance will be affected by the parameters of the model. However, how to effect diagnose and classify disease using the optimum model is becoming an urgent issue. In this paper, we propose two new classifiers that can automatically search for the optimum parameters of the model. We called these two classifiers are Nested–Random Forest (Nested-RF) classifier and Nested–Support Vector Machine (Nested-SVM) classifier. Five datasets of cancer (brain cancer, colon cancer, DLBCL, leukemia, prostate cancer) and one disease (Parkinson's) datasets were used to evaluate the performance of the proposed classifiers. Our results show the superior performance of the Nested-SVM classifier. Compared to the other three classifiers, the Nested-SVM classifier can improve classification performance (ranged from 2 to 5% in accuracy, sensitivity, and specificity) in cancer classification. In Parkinson's disease classification, the Nested-SVM classifier shows the superior performance with the accuracy up to 93% that are 20% more than the results from other three classifiers. The results imply that the Nested-SVM classifier has the potential of becoming the standard of setting classifier parameters and maybe suitable for the diagnosis of patients with cancers and Parkinson’s disease.