Ischemic stroke remains a significant cause of mortality and morbidity globally, necessitating accurate predictive models for early diagnosis and intervention. This study proposes an enhanced approach by integrating advanced machine learning techniques to predict and interpret ischemic stroke outcomes. The existing system employs Convolutional Neural Networks (CNN), Long ShortTerm Memory networks (LSTM), and XGBoost for predictive analytics. However, to overcome limitations and enhance performance, our proposed system introduces the Passive Aggressive Classifier (PAC), Decision Tree, and Adaboost Classifier into the model concatenation framework. The proposed framework leverages the strengths of each model type: PAC for its efficiency in handling largescale data streams with limited computational resources, Decision Tree for its interpretability and robust handling of feature interactions, and Adaboost for its ability to improve weak learners sequentially, thereby boosting overall predictive accuracy. By integrating these diverse models, our approach aims not only to enhance predictive performance but also to provide interpretable insights into the factors influencing ischemic stroke occurrence and severity.

Keywords: Ischemic Stroke Prediction, Deep Learning Models, Machine Learning Fusion, Passive Aggressive Classifier, Decision Tree, Adaboost Classifier, Predictive Analytics, Interpretability.