Machine Learning for Real-Time Cardiovascular Disease Prediction Based on Cloud

Abstract

   Diagnosing cardiovascular disease is an essential medical process to guarantee accurate classification, which aids cardiologists in treating patients appropriately. By employing machine learning for the cardiovascular illness classification, it is possible to decrease the occurrence of misdiagnosis and save patients' lives. This study has developed an effective Amazon Web Services (AWS) machine learning architecture model to predict cardiovascular diseases. It integrates a number of AWS services, including SageMaker, Lambda, API Gateway, and S3, which provide significant automation and real-time prediction. The cardiovascular dataset from Kaggle is classified using the ensemble algorithms, including CatBoost, XGBoost, and LightGBM. To improve classification accuracy, k-modes clustering with Huang initialization is used. Additionally, SageMaker Automatic Model Tuning is utilized to optimize a model's hyperparameters based on Bayesian optimization. The experimental results indicate that the CatBoost classifier outperformed with a slightly higher accuracy of 87.9% compared to the other applied algorithms. Furthermore, the model takes about 199 ms to respond to the prediction request, which makes it quick and appropriate for applications that require low latency.