Evaluation of machine learning models for predicting heart disease

Heart disease is one of the most dangerous diseases that can affect humans and leads to a high mortality rate, so we need to develop effective prediction systems. This paper presents a comprehensive analysis using many machine learning methods for heart disease forecasting based on "Cleveland Heart Disease" dataset. This dataset contains 303 patient "records" with 14 "attributes", such as age, gender, blood pressure, cholesterol levels, and other potential risk factors for heart disease. The algorithms that used are: Naive Bayes, Random Forest, Gradient Boosting, Support Vector Machine (SVM), XGBoost, CatBoost, and LightGBM. The performance depended on many aspects which includes: accuracy, precision, recall, F1 score, and AUC-ROC. The results illustrate that LightGBM gets the highest degree of 83.5%, in other hand Gradient Boosting gets the better AUC-ROC of 0.89. Naive Bayes and Random Forest achieved a performance score of 82.4%, with Random Forest achieving the best values among the methods used with fine-grained variables. Furthermore, XGBoost and CatBoost executed admirably, via CatBoost obtaining 79.1% accurate and XGBoost reaching 81.3%. Voting The taken-together categorization of XGBoost, CatBoost, and LightGBM achieved 81.3% efficiency, highlighting the effectiveness of ensembles approaches. The novelty of the paper is giving a full comparison examination of a range of approaches to machine learning, including cutting-edge collective approaches such as XGBoost, CatBoost, and LightGBM, in connection with predicting heart diseases. The study emphasizes the advantages of using these advanced algorithms in medical diagnostics and provides useful information about how they can be used and efficacy in real-life situations.