Heart disease remains one of the primary causes of mortality globally, emphasizing the need for accurate early detection systems. This paper presents a comprehensive framework for predicting the risk of heart disease using advanced machine learning techniques. We propose hybrid models that combine several state-of-the-art classification algorithms to improve prediction accuracy. Specifically, we evaluate and compare models such as Stacked Autoencoder with Random Forest, a Hybrid model combining XGBoost and LightGBM, the Stacking Classifier, and the Voting Classifier. These models are trained on a dataset sourced from Kaggle, which includes key features such as age, cholesterol levels, blood pressure, and medical history. The hybridization of these algorithms leverages their individual strengths to enhance the performance of the prediction system. By stacking models and using voting mechanisms, the proposed approach aims to provide more reliable and precise predictions. The results demonstrate that hybrid models outperform traditional single-model approaches, offering significant promise in the early detection of heart disease. This research contributes to the ongoing efforts in utilizing machine learning for healthcare, specifically in reducing the mortality rate associated with heart disease through timely and accurate predictions.

Keywords: Heart disease prediction, Machine learning, Stacked Autoencoder, Random Forest, XGBoost, LightGBM, Stacking Classifier, Voting Classifier, Early detection, Hybrid models, Healthcare, Kaggle dataset.