Combining Deep Learning and Machine Learning Techniques to Predict Cardiovascular Disease Accurately from Electronic Health Records

Identifying high-risk patients early on is crucial to improving treatment outcomes since one of the leading causes of mortality globally is cardiovascular disease (CVD). Using electronic health records (EHRs), this study investigates how well different machine learning and deep learning algorithms predict CVD. Using a sizable EHR dataset, we assess the effectiveness of many ensemble techniques, such as Gradient Boosting, Cat Boost, Random Forest, Ada Boost, and Light GBM (LGBM), as well as deep learning models such as TABNET and Long Short-Term Memory (LSTM). Our findings show that TABNET has the best accuracy among the evaluated deep learning models, whereas Cat Boost performs better than other ensemble approaches. In addition, we look at the significance of features and how hyper parameter tweaking affects model performance, offering guidance on how to maximize predictive power. This study highlights how sophisticated algorithms can improve CVD early detection and aid in clinical decision-making.