A Comprehensive Analysis of Machine Learning Models for the Prediction of Diabetes Mellitus

This survey paper provides an overview of the current state-of-the-art techniques for diabetes diagnosis and prediction using machine learning (ML) methods. Various techniques, including both basic classifiers and deep learning networks, have been applied to solve the PIMA Indian Diabetes Dataset (PIDD) and other similar datasets. These include diagnostic methods that employ deep learning architectures such as interpretable filter convolutional neural networks (IF-CNN), CNN-Bi-Directional LSTM hybrids, and grasshopper optimization algorithms. These approaches have led to improvements in prediction accuracy, often through feature selection techniques such as Principal Component Analysis (PCA) and Polynomial Regression (PR). However, many challenges persist. For example, the large amount of data required to train Spiking Neural Networks (SNNs) is a limitation, while the models are often less robust to noise, with no effective way to monitor network performance. This survey aims to compare these models, highlight their strengths and weaknesses, and provide insights for future advancements in predicting diabetes mellitus.