Abstract

Cardiovascular disease has been the major cause of death in the past years. According to the world health organisation statistical analysis in 2021, 17.9 million people die of heart disease annually (WHO 2021). It has been a major concern that most medical practitioners treat patients based on their knowledge and intuition, which is not good enough as so many lives have been lost based on the error in diagnoses. The medical field has shown interest and vast development in using machine learning to solve some medical problems by using patients’ medical records to create solutions to some of the problems arising through the use of artificial intelligence.

In the proposed study, an ensemble learning technique has been proposed for developing cardiovascular disease prediction. The research study utilised a combined dataset from four database repositories, Hungary, Switzerland, Cleveland, and Long Beach V. The study is a supervised classification problem that is based on predicting if a person has a disease or not. Several algorithms were implemented for performance comparison which are Logistic Regression, Random Forest, Decision Tree, KNN, Multilayer Perceptron, Extreme Gradient Boost and Hybrid model (Hybrid of Random Forest, Decision Tree, and Extreme Gradient Boosting).

The models used for ensemble techniques were chosen based on the best three models with good performances. Models were tuned to improve the performances using manual search and random search techniques for the best parameter settings. Model evaluation was carried out by accuracy of models, precisions, F1-score, recall, confusion matrix and ROC curve. A cross-validation score was also used for the estimation the performance of learning models. All models gave a good result, but the ensemble method was the best with 99% accuracy. A friendly graphical user interface was developed for easy access to users. The GUI was designed based on the open-source framework known as stream lit, with the ensemble model of Extreme Gradient Boosting, Random Forest, and Decision Tree to predict.

Heart Disease Classification Based on Machine Learning and Deep Learning