Development and Performance Evaluation of a Heart Disease Prediction Model Using Convolutional Neural Network

Abstract

Heart disease is a leading cause of mortality globally and its prevalence is increasing year after year. Recent statistics from
the World Health Organization show that about 17.9 million individuals are embattled with heart diseases annually and people under
the age of 70 account for one-third of these deaths. Hence, there is need to intensify research on early heart disease prediction and
artificial intelligence-based heart disease prediction systems. Previous heart disease prediction systems using machine learning
techniques are unable to manage large amount of data, resulting in poor prediction accuracy. Hence, this research employs
Convolutional Neural Networks, a deep learning approach for prediction of heart diseases. The dataset for training and testing the
model was obtained from a government owned hospital in Nigeria and Kaggle. The resulting system was evaluated using precision,
recall, f1-score and accuracy metrics. The results obtained are: 0.94, 0.95, 0.95 and 0.95 for precision, recall, f1-score and accuracy
respectively. This show that the CNN-based model responded very well to the prediction of heart diseases for both negative and positive
classes. The results obtained were also compared to some selected machine-learning models like Random Forest, Naïve Bayes, KNN
and Logistic Regression and results show that the developed model achieved a significant improvement over the methods considered.
Therefore, convolutional neural network is more suitable for heart disease prediction than some state-of-the-art machine-learning
models. The contribution to knowledge of this research is the use of Afrocentric dataset for heart disease prediction. Future research
should consider increasing the data size for model training to achieve improved accuracy