Stroke remains one of the foremost causes of mortality and long-term disability worldwide. Early prediction of stroke is crucial in enabling timely medical interventions and minimizing its adverse outcomes. This study presents a robust ensemble- based machine learning framework that integrates the predictive strengths of XGBoost, a Multi-Layer Perceptron (MLP), and a hybrid Bidirectional LSTM-GRU neural network to enhance the accuracy of stroke prediction. The model is trained and validated using a publicly available healthcare dataset containing a diverse set of clinical and demographic features such as age, hypertension, heart disease, smoking status, and body mass index (BMI). Comprehensive data preprocessing steps, including missing value imputation, feature encoding, and class balancing using Random Over Sampler, were implemented to prepare the dataset for training. Each individual model was optimized for performance and then combined in an ensemble using weighted averaging to maximize generalization and robustness. The pro- posed ensemble model demonstrated superior performance with an accuracy of 94.58%, precision of 90.64%, recall of 99.45%, F1- score of 94.84%, and an impressive ROC-AUC score of 98.73%. The high recall indicates that the model is particularly effective in identifying stroke-positive cases, which is critical in a clinical setting. The confusion matrix and classification report further support the model’s reliability and effectiveness. Additionally, visualization techniques such as ROC curves, precision-recall curves, and calibration plots confirmed the model’s strong predictive confidence. Feature importance analysis highlighted the significant impact of variables such as age, hypertension, and BMI. This ensemble framework presents a powerful tool for clinical decision support, with potential deployment in real- time health monitoring systems for early stroke detection and intervention.
This research presents a novel architectural paradigm for digital image and video processing systems...
Encryption makes it possible to send private information via unprotected channels without worrying a...
Kidney diseases such as cysts, stones, and tumors are among the most prevalent health issues worldwi...
This project is an interactive job and internship portal developed using the MERN stack with custom ...
