The Role of a Combined Multi-Layer Perceptron Classifier in Achieving Superior Diabetes Prediction

Being a chronic disease, Diabetes mellitus is very common throughout the whole world and continues to exert inordinate pressure on healthcare systems and economies of various countries. Timely intervention and diagnosis of at-risk individuals can avoid serious complications. Diabetes mellitus is prevalent the world over as a chronic illness; therefore, it continues to be an expense to health care systems and economies of various countries. Early detection of persons at risk should be before they suffer serious complications so that immediate medical care may be given to them. In this study we evaluate all algorithms including Random Forest (RF), K-Nearest Neighbor (KNN), Support Vector Machine (SVM), Multilayer Perceptron (MLP), XGBoost, Convolutional Neural Networks (CNN), and Long Short-Term Memory (LSTM). An MLP came first among all others with 98.3 % accuracy, 100 % precision, and 98.4 % F1 score. The SVM, RF, and XGBoost all had very good predictions also while the CN, KNN, and LSTM were acceptable with slightly lower prediction values.