An Innovative Machine Learning Pipeline for Stroke Prediction on Imbalanced Data

Stroke requires immediate medical care. The WHO states stroke is the leading cause of death and disability world-wide. Thus, early stroke detection is essential. Few studies projected with CNNs or other stacking approaches. CNN-ensemble classifier. This study used unequal Kaggle data. Random over-sampling balances data. After identifying the most important features, several machine learning algorithms are applied, including Logistic Regression, Support Vector Machine, Decision Tree Classifier, Random Forest Classifier, K-Nearest Neighbor's Classifier, Gaussian Naive Bayes, Bernoulli Naive Bayes, and Stacking for Six Algorithms (Support Vector Machine, Decision Tree Classifier, Random Forest Classifier, KNN, BernoulliNB, GaussianNB). Training and testing evaluate each algorithm's stroke prediction accuracy. Stacking six algorithms gives the maximum accuracy, testing is
$\mathbf{99.9}\pm\%$
and training is 100%. Machine learning improves this classifier's stroke prediction.