A Method for Classification Using Data Mining Technique for Diabetes: A Study of Health Care Information System

Many researchers in the health information system field have been attracted to develop computer applications that help in the diagnosis process. Imperatively, data mining algorithms address the vital role in all of these applications. Many contributions were made in this area. There has always been a debate on the algorithm that gives the best classifier, the parameters to be used, the dataset pre-processing steps, etc. In this paper, the author largely emphasizes that the best way to build a predictive model with relatively high classification accuracy is to build several predictive models and to choose the model that gives the best results through parameters optimization. Diagnosing diabetes mellitus has gained considerable attention in the last few decades due to the increased severity of the disease. In this research, the author reviews four predictive data mining approaches that are being used in diagnosing diabetes. Four models were implemented to diagnose diabetes from PIMA dataset; k-nearest neighbour, support vector machine, multilayer perceptron neural network, and naive bayesian network. Giving the highest classification accuracy, support vector machine technique outperformed the others with a value of 78.83%.