Early Prediction of Diabetes using Machine Learning and Deep Learning Techniques – A Comparative Analysis

Abstract

One of the most common chronic illnesses in the world, diabetes mellitus has grown in
frequency and complications, making it a serious public health issue. Early detection and
treatment of diabetes can enhance healthcare management and drastically lower death rates.
Recently, methods for machine learning (ML) and deep learning (DL) have demonstrated
encouraging results in disease prediction systems. Using the PIMA Indians Diabetes Dataset,
this work compares a few of machine learning and deep learning methods for diabetes
prediction. The suggested comparative framework assesses machine learning techniques like
Logistic Regression (LR), Support Vector Machine (SVM), and Extreme Gradient Boosting
(XGBoost) and deep learning techniques like Convolutional Neural Network (CNN), Long
Short-Term Memory (LSTM), and Bidirectional Long Short-Term Memory (BiLSTM).
Results from experiments show that deep learning models perform better in prediction than
traditional ML approaches. Our analysis found that the BiLSTM model performed about
98.7% better than the other algorithms.