An Experimental Study of Various Optimization Techniques for Genetic Disorder based diseases Prediction Using Deep Learning Techniques

Abstract

Genetic disorders are the result of any changes or defects in genes or chromosomes. Detecting those diseases in an early stage leads to the treatment process make easier. In recent years, the advancement in Big data field and deep learning techniques have created new opportunities for predicting these disorders in an early stage itself. Deep learning models can identify complex patterns in genetic data, but their effectiveness enhanced by using optimization methods applied during the training process. The study evaluates the performance of Aquila Optimizer (AO), Enhanced Pelican Optimization (EPO), Modified Gannet Optimization (MGO), Sparrow Search Algorithm (SSA), AdamBoost, and Gradient Boost algorithms integrated with the deep learning models such as Temporal Convolution Network, Bi-Directional LSTM and Self Attention mechanism. The study analyzes these optimization techniques based on accuracy, convergence speed, loss minimization, and computational efficiency. Experimental results demonstrate that bio-inspired and hybrid optimizers significantly improve prediction accuracy and convergence behavior compared to traditional gradient-based approaches.