Abstract
Now-a-days, diabetes disease is considered one of the key reasons of death among the people in the world. The availability of extensive medical information leads to the search for proper tools to support physicians to diagnose diabetes disease accurately. This research aimed at improving the diagnostic accuracy and reducing diagnostic miss-classification based on the extracted significant diabetes features. Feature selection is critical to the superiority of classifiers founded through knowledge discovery approaches, thereby solving the classification problems relating to diabetes patients. This study proposed an integration approach between the SVM technique and K-means clustering algorithms to diagnose diabetes disease. Experimental results achieved high accuracy for differentiating the hidden patterns of the Diabetic and Nondiabetic patients compared with the modern diagnosis methods in term of the performance measure. The T-test statistical method obtained significant improvement results based on K-SVM technique when tested on the UCI Pima Indian standard dataset.