In this paper, an enhanced Support Vector Machine framework, called Improved Support Vector Regression (ISVM), has been proposed for classification applications in healthcare datasets. The current traditional classifiers showed inability to handle data with a variety of data structures, rather small sample sizes, and nonlinear relationships that are often found in complex fields like medical applications. The performance and the prediction power of the proposed approach were assessed by comparing it with the employed machine learning techniques of SVM, Logistic and Naïve Bayes. Accuracy, sensitivity and specificity were used to measure performance. Results show that the ISVR model outperforms the benchmark models in most scenarios. ISVR demonstrated better stability in noisy settings and higher predictive accuracy with larger sample sizes, thus indicating high generalization ability. ISVR had the best classification accuracy with lower variance than SVM and LR, and NB had lower accuracy as it made strong independence assumptions