In this study, the healthcare industry relies on medical time-series analysis to facilitate precise diagnosis, patient monitoring, and malady prediction. Nevertheless, conventional machine learning models frequently encounter obstacles due to the intricate, sequential character of medical data, including electrocardiograms (ECG), electroencephalograms (EEG), and glucose
level fluctuations. Evolutionary Algorithms (EAs), including Genetic Algorithms (GA), Particle Swarm Optimization (PSO), and Differential Evolution (DE), are increasingly employed
to enhance feature selection, optimize predictive models, and adjust hyperparameters for increased accuracy in order to overcome these obstacles. This study investigates the feasibility of integrating EAs with deep learning models, such as Transformer-based architectures, Gated Recurrent Units (GRU), and Long Short-Term Memory (LSTM), to improve prediction
performance in medical time-series analysis. The study ensures the development of a dependable model by employing publicly accessible datasets, such as the PhysioNet ECG Database, MIMIC-III Waveform Database, and UCI Epileptic Seizure Dataset. EAs enhance forecasting accuracy in a hybrid manner by adjusting model hyperparameters, such as
the learning rate, dropout rate, and sample size. The Transformer + DE model achieved the maximum accuracy (91.2%) outperforming baseline LSTM and GRU models, as evidenced by the testing results, which indicate that EA-optimized models outperform traditional deep learning Additionally, the efficacy of EA-driven tailoring is underscored by optimization convergence experiments and feature importance analyses. Evolutionary optimization is a potent approach to medical time-series forecasting that enhances the identification of anomalies, the modeling of disease progression, and the surveillance of patients in real-time, despite its processing challenges. This work underscores the potential of integrating EAs with artificial intelligence to facilitate personalized healthcare, early disease identification, and improved clinical decision-making, thereby facilitating future advancements in medical
informatics.