Machine intelligence in stroke prediction

Abstract

The innovative developments in the field of machine intelligence have paved way to the growth of tools for assisting physicians in disease diagnosis. Early diagnosis and prognosis of stroke are crucial for timely prevention and cure. This research work focuses on the design of a stroke prediction system by investigating the various physiological parameters that are used as risk factors. Features extracted from various risk parameters carry vital information for the prediction of stroke. Classification algorithm that has been used with the number of attributes for prediction are support vector machines (SVM) and artificial neural network. Data collected from international stroke trial database was successfully trained and tested using both classifiers. The predictive models discussed here are based on different supervised machine learning techniques as well as on different input features and data samples. SVM gave an accuracy of 91% while neural network outperforms SVM by providing an accuracy of 98.1%.