Predicting the Risk of Diabetes Using Explainable Artificial Intelligence

Diabetes is a metabolic condition depicted by elevated blood glucose levels where the pancreas is unable to generate adequate insulin or the body is incapable of utilizing insulin properly, thus incapable to transport the glucose into the cells for energy. Prompt diagnosis and treatment, more particularly lifestyle modification, can avert or delay the emergence of microvascular and macrovascular complications. The aim of the study is to predict the risk of developing diabetes using a variety of machine learning classification algorithms applied on a publicly available diabetes dataset as well as to investigate the impact of the feature selection techniques and class imbalance handling method on the predictive power of these models The model with the best performance was achieved by random forest with synthetic minority oversampling method and without any feature selection technique with an area under the curve of 0.983, 99% accuracy, 100% recall, 100% precision and 100%. Moreover, a simple web-based application that predicts risk of acquiring diabetes was also developed as a decision support tool. It was also integrated with explainable artificial intelligence method to highlight the characteristics leading towards the prediction decision of the model. Prompt and accurate diabetes recognition can assist patients to foresightedly incorporate lifestyle habit modifications so as to ward off risks of complications of uncontrolled diabetes.