The SARS-CoV-2 virus has presented unparalleled global health challenges through its high contagiousness and substantial mortality rates in the COVID-19 pandemic. Amidst strained community health services, there's a pressing demand to identify mortality predictors and enhance patient care. This study introduces an innovative approach, utilizing ensemble learning techniques for COVID-19 mortality prognosis based on accessible blood test outcomes. By scrutinizing diverse blood parameters—age, LDH, lymphocytes, neutrophils, and hs-CRP—five influential markers are pinpointed. Their amalgamation yields impressive precision, accurately forecasting mortality in 96% of instances. Through a fusion of XGBoost feature importance and neural network classification, the method attains exceptional predictive prowess, maintaining a 90% precision up to 16 days prior. Rigorous validation across distinct scenarios validates the model's durability and applicability. These biomarker insights furnish practical insights and avenues for swift decision-making in tailored medical interventions. With its prompt, accurate, and dependable methodology, this study holds the potential to transform the field, guiding effective strategies in the ongoing battle against COVID-19. Our study showcases the power of ensemble learning and rapid data processing in identifying crucial biomarkers for predicting COVID-19 mortality. The method's high precision and early predictive capability offer a promising tool for healthcare practitioners and policymakers to make informed decisions and allocate resources effectively in the fight against COVID-19.