Optimal multi-layer perceptron parameters for early stage diabetes risk prediction

Abstract

Abstract Diabetes is an alarming threat to people around the world because the number of diabetics is increasing every year. Diabetics with other complications have a very high risk of death. Therefore, the use of technology to predict the risk of early diabetes is needed. Neural Network as one part of artificial intelligence plays a role in solving prediction problems with satisfying results. In this study, a multi-layer perceptron neural network is used to predict the risk of early stage diabetes with optimal parameters from the optimization results using Improved Crow Search Algorithm. The test results prove that the multi-layer perceptron with optimal parameters is able to provide better accuracy compared to other algorithms such as J48, PART, Decision Table, Naïve Bayes, AIRS1, AIRS2, and Single Layer Perceptron with the highest accuracy values of 97.69% and 96.92% for one and two hidden layers, respectively. This proves that the proposed solution can be used to predict the early stage diabetes risk.