Efficient Machine Learning-Based Model for Cerebrovascular Disease Prediction with Nigerian Dataset

Abstract

Cerebrovascular accidents (CVA) or stroke has been a global phenomenon that causes disability and deaths of people worldwide, particularly in the middle- and low-income countries. It has been reported that more than 100,000 cases are recorded every year in Nigeria. Moreover, several deaths were reported globally by the World Health Organization (WHO). Diagnostic tools, preventive measures, and medical experts are insufficient and contribute to the escalation of the disease worldwide. Several predictive models have been proposed by scholars but have been inadequate due to variability in the risk factors, race, and geographical variations. This paper compared six machine learning-based models with three feature selection algorithms on a Nigerian dataset containing 103 instances with 22 features. We trained and evaluated the NB, SVM, LR, MLP, J-48, and RF with CBFS, CAE FS, and Relief FS algorithms. The results of our experiments showed that the J-48 model with the CBFS algorithm was computationally faster and achieved an excellent prediction accuracy of 100.00% in 0.00 seconds. The type of data used has a substantial impact on the performance of machine learning classifiers. Therefore, based on the experiments performed, J-48 with CBFS algorithm was proposed for deployment as the clinical decision support system that could assist medical professionals in predicting cerebrovascular diseases.