Cerebral Stroke Classification Using Over Sampling Technique and Machine Learning Models

In recent years, cerebral stroke has ascended as a paramount concern in global public health. Proactive strategies emphasizing metabolic control over salient risk factors present a superior approach compared to relying solely on physiological indicators, which may not delineate clear preventive directives. In this research, we present the SPX-CerebroPredict model—a novel machine learning framework designed to classify imbalanced cerebral stroke data for clinical diagnostics. The study delves into feature selection methodologies, employing both information gain and principal component analysis (PCA). To address the class imbalance dilemma, the Synthetic Minority Over-sampling Technique (SMOTE) was harnessed. The empirical evaluation, conducted on the cerebral stroke prediction dataset from Kaggle—comprising 43,400 medical records with 783 stroke instances—pitted well-established algorithms such as support vector machine, logistic regression, decision tree, random forest, XGBoost, and K-nearest neighbor against one another. The results evince that our SPX-CerebroPredict model, integrating SMOTE, PCA, and XGBoost, surpasses its contemporaries, achieving an impressive accuracy rate of 95%. This discovery underscores the model’s potential for clinical applicability in cerebral stroke diagnostics.