As per World Health Organization (WHO) Brain stroke is the second underlying cause of death categories or major ICD(International Cause of Death) and disability across the globe . Artificial Intelligence (AI) enabled approaches using Machine Learning (ML) are widely used for stroke detection automatically in a non-invasive fashion with data-driven approach. However, from the literature, it is understood that there is need for improving quality of training and also find best classifiers to ensemble them to enhance prediction performance. In this paper, we proposed a framework known as Stroke Prediction Ensemble (SPE) which exploits a hybrid approach considering feature engineering and ensemble classification. From multiple brain stroke prediction models, best models that exhibit accuracy >90% are chosen for ensemble model. Two algorithms are proposed to realize the framework. They are known as “Hybrid Measures Approach for Feature Engineering (HMA-FE)” and Hybrid Ensemble and Feature Engineering for Stroke Prediction (HEFE-SP). The former is published in our prior work which is meant for finding best features from given dataset while the latter is meant for ensemble ML towards more efficient stroke prediction performance. Empirical study has revealed that our ensemble model showed highest accuracy with 97.93% while the average accuracy of all constituent base line models is 95.25%. Thus the ensemble model can be used for efficient brain stroke diagnosis as part of Clinical Decision Support System (CDSS).