Evolutionary algorithm-based hyperparameter tuning of one-dimensional CNNs for diabetes mellitus prediction

Diabetes mellitus, a pervasive and intricate metabolic disorder, disrupts hormonal equilibrium, resulting in elevated glucose levels and widespread organ impact. Timely and precise diagnosis is pivotal for effective disease control. This research introduces an advanced Deep 1D-Convolutional Neural Network (1DCNN) tailored for diabetes classification, specifically addressing challenges posed by imbalanced datasets and missing values. Through three experiments, encompassing a baseline 1DCNN, hyperparameter optimization(HO) using a Genetic Algorithm (GA), and a Particle Swarm Optimization (PSO) comparison, we aim to identify the most effective model configuration. Our study yields improved results compared to the state-of-the-art. Achieving high precision, recall, and accuracy, along with superior AUC and PR curves, our research significantly contributes to the refinement of diabetes prediction models, fostering enhanced disease comprehension and control.