Early Detection of Stroke Risk Using Optimized Light Gradient Boosting Machine Approach Based on Demographic Data

Abstract

Stroke is a clinical condition wherein blood vessels inside the brain rupture, resulting in brain damage. Symptoms may appear if the brain's blood flow and other nutrients are disrupted. Early identification of different stroke warning signals can assist in lessening the severity of the stroke. This research study identifies early stroke diseases by utilizing an ensemble learning strategy using clinical data such as BMI, hypertension, average glucose level, heart disease, smoking status, and other factors of several individuals from Bangladesh. This study suggests utilizing the light gradient boosting machine (LGBM), an ensemble learning technique, to identify stroke risk prediction, with the data resampled and the parameters modified using grid-search. The proposed method's performance has been validated using a variety of machine learning algorithms, including DT, KNN, MLP, LDA, LR, SGD, GNB, and QDA. The experimental findings revealed that the proposed model surpasses all other performance measures such as precision, accuracy, TPR, TNR, FPR, F1-score, and AUC-ROC.