Diagnosis of Heart Stroke Using Feature Extraction and Sequential Learning Based Models

In today's world, Heart Failure and Stroke are one of the leading causes of death worldwide and are the third leading cause of death and long-term illness in the United States. As per American Heart Association, there are about 405 deaths from stroke each day which implies that every 3 minutes, someone dies of a stroke. Accurate prediction of such diseases is highly valuable for early medication and further treatment. Our study revolves around comparing various classifier models in machine learning for predicting heart stroke/failure at an early stage depending upon the conditions of the patient. The experiment comprises the evaluation part including different resultant parameters namely accuracy, precision, recall, sensitivity, and F1-score. In the experiment, we have used the dual dataset approach for better analysis. The datasets has been selected keeping in mind the key purpose of having as many as possible different attributes that help in the prediction of the stroke. The experiment comprises 11 classifier models comparison on each dataset. The experiment results show that Random Forest and Artificial Neural Network classifier shows the best results having 86.66% and 86.84% of accuracy for two datasets respectively. The results also showed that the XGBoost classifier model can also be used because of its high recall factor.