Heart Attack Prediction uisng Random Forest

Cardiovascular illnesses and cardiac arrests, in particular, have become the major cause of death allover the world, highlighting the need to develop new methods for early prediction and diagnosis for better treatment and prevention. The proposed work integrates the application of Random Forests in a predictive ML (ML) model boosted with Explainable Artificial Intelligence (XAI) techniques, to be able to predict if the cardiac arrest could happen or not. The RF model was trained using a clinical database consisting of demographic, social and medical data and was able to predict the chances of a cardiac arrest with an increased level of precision. XAI techniques like SHAP (Shapley additive explanations) and Partial dependency plots (PDP) were utilized to improve the accuracy and the reliability of the model with XAI. Together, these techniques helped to explicate the forecast yielded by the model as to what features were important and what were the predictions that led towards the conclusions in the model. Cholesterol levels, along with blood pressure and age, were among the key indicators scrutinized that supported the functioning of the model. The results presented in the last section highlighted the main risk factors and provided additional relevant information to healthcare providers, which could greatly improve their decision-making and treatment of patients by knowing what treatments to provide to which patients. The investigation addresses the challenge of the medical AI paradigm by overcoming the trade-off between prediction performance and interpretability by developing AI devices that are clinically usable from the use of ML to XAI. The proposed approach enables the prediction of cardiac arrests at an early stage, with the goal of assisting patients and minimizing the costs incurred by the health facilities, by increasing the anticipatory, allowing improved patient care and preventive methods for patients.