Brain Stroke Prediction by Machine Learning Algorithm Along with Boosting Classifier

A stroke is a medical emergency when blood circulation in the brain is disrupted or outflowing due to a burst of nerve tissue. Diagnosis at the proper time is crucial to saving lives through immediate treatment. Machine learning techniques show good accuracy in predicting the likelihood of a stroke from related factors. Ten machine learning classifiers have been considered to predict stroke namely, logistic regression (LR), Random forest (RF), Decision tree (DT), Support vector machine (SVM), K-nearest neighbors (KNN), Gaussian Naïve Bayes (GNB), and four boosting classifiers which are Gradient boost, adaboost, Xgboost and lightgbm. SMOTE is applied to balance the imbalance dataset and then again used in all the models. It is found that Random forest has the highest 96% accuracy, extreme boost and lightgbm both have 95% accuracy. Here traditional logistic regression and Gaussian Naïve Bayes have higher error, which is 21% and 24%.