Coronary Artery Disease Diagnosis with Deep Neural Network, Lightgbm and XGBoost

Document Type : Research Paper


1 Mechanical Engineering department, Islamic Azad University of Tehran South Branch

2 Faculty of New Sciences and Technologies, University of Tehran, Iran

3 Artificial Intelligence department, University of Isfahan, Iran



Background and Objectives: Artificial intelligence and machine learning methods have proved to be able to solve both data analysis and classification problems in many fields like medical diagnoses. With development of technology in many areas like processing units and waste memory storage in recent years, many new approaches have come into reality from prolepses such as deep neural networks and gradient boosting machines. These new models are now able to classify any type of data with high precision and accuracy. They are also able to face many challenges, including imbalance data and nonlinear dependencies in high dimensional spaces. These abilities make new methods a lot more reliable and popular.
Methods: In this study, an imbalance medical dataset is used to detect heart disease by ensembling three different models including deep neural networks (DNN), light gradient boosting machine (LightGBM) and XGBoost.
Results: As implementation results show, these methods are effective and robust while they reach an accuracy of 91.75% and f1_score 94.4.
Conclusion: In this study, an imbalance medical data set is classified using an ensemble method to diagnose heart disease with high accuracy.