Predicting the occurrence of Ischemic stroke by Gradient Boost Approaches

If we look around the globe then the leading cause of mortality and disability can be attributed to Ischemic strokes. Early detection and prediction of stroke occurrence can lead to better treatment outcomes and improved patient care. Wit the advent of technology, machine learning approaches have been increasingly applied to stroke prediction and diagnosis. Gradient Boosting is a popular ensemble classification technique that can be used to construct accurate predictive models for various applications, including stroke prediction. In this study, we compare the various Gradient Boosting approaches to accurately classify people which might have ischemic stroke and measure the efficiency of each of these algorithms. The various algorithms namely the XGBoost, LightGBM and CatBoost are compared by tackling the various problems of data preprocessing like class imbalance. The algorithms are compared by tuning the hyperparameters and the results show that they can accurately predict the occurrence of ischemic stroke with an accuracy of about 96%. The feature importance analysis reveals that age, hypertension, and BMI are the most significant predictors of stroke occurrence. The results conclude that XGBoost overperforms the other two algorithm and can be a promising approach to detect and determine the ischemic strokes, allowing for better handling of the medical facilities and timely prevention of the disease improving patient healthcare.