Prediction of Diabetes Using Data Mining Techniques

Diabetes mellitus is fourth most high mortality rate diseases in the world and it is also a cause of kidney disease, blindness, and heart diseases. Data mining techniques support a medical decision for a correct diagnosis, treatment of disease in such way it minimizes the workload of specialists. This study proposed to predict diabetes using data mining techniques. Back propagation algorithm is used to predict whether the person has diabetic or not. And also J48, naive bayes and support vector machine were used to predict diabetes. These neural networks were having an input layer with having 8 parameters, one hidden layer having 6 neurons and produce one output layer.5 fold cross-validation technique and large value learning rate was used to improve the performance of the model. PIMA Indian dataset used to conduct this study. The study implemented in RStudio using R programming language. The performance of Back propagation algorithm is used to predict diabetes diseases gave 83.11 % accuracy, 86.53% sensitivity and 76% specificity, the result shows improvement from previous work. The obtained result is also compared with J48, naive bayes and support vector machine algorithm.