A Deep Learning Approach for Type-2 Diabetes Risk Prediction – A Review

Searching the vicinity of the fuzzy boundaries, type 2 diabetes is one of the most widespread chronic diseases that demand early identification for its effective treatment and prevention from severe complications. The advent of machine learning (ML) and deep learning (DL) as predictive healthcare techniques has put forward the opportunity for superior accuracy than traditional statistical approaches. This survey paper endeavors to include various algorithms used in ML and DL for prediction in Type 2 diabetes, their methods of operation, and broad performance analysis with challenges encountered therein. A survey of some of the ML models such as K-Nearest Neighbors (KNN), Random Forest, Support Vector Machines (SVM), Decision Trees, Logistic Regression, and ensemble methods along with the DL techniques like Recurrent Neural Networks (RNNs), Convolution Neural Networks (CNNs), Artificial Neural Networks (ANNs), Long Short-Term Memory (LSTM) is presented in this paper. This study, further, discusses the importance of feature extraction, pre-processing, hyper-parameter tuning, and regularization techniques that act as add-ons to bolster one's performance. We also review benchmark datasets the Pima Indians Dataset and the datasets from clinical studies applicable to diabetes prediction. The paper concludes with observations on current challenges, future research directions, and the incorporation of explainable AI (XAI) techniques for the enhancement of interpretability and clinical applicability of the model.