A Comprehensive Review of Diabetes Mellitus Exposure and Prediction using Deep Learning Techniques

Nowadays Diabetes disease is currently most prevalent ailments in the world. Due to recent technical breakthroughs, Machine Learning approaches have helped significantly to the disease prediction process. This has motivated researchers to investigate the difficulty of this problem using Deep Learning (DL) algorithm sets. A CNN-LSTM combination model has attained the highest accuracy of 95.1% to date. Although several methods were employed to tackle this problem using this technique, there are specific sets of classifiers that are either rarely used or not used at all, based on their predictions. In addition, no current survey sources have evaluated and equated the recital of ML as well as DL in coupled models. The determination of this article is to afford an outline of all diabetes forecasts made using ML and DL approaches over the past six years. Also, a recent study on the Pima Indian Dataset aimed at implementing those unused ML classifiers and analyzing their performances. They acquired an accuracy of about 68% to 74%. By this way, we can enhance these classifiers in developing combined models and also in diabetes prediction.