A Novel Hybrid Approach Integrating Machine Learning and ANN for Stroke Risk Assessment

Brain stroke is a life-threatening condition that needs to be detected early and correct precision to improve patient wellness. This paper presents a method for predicting stroke using ML and ANN using a publicly available data from Kaggle. The dataset has class imbalance. This can be solved using SMOTE and ADASYN to guarantee a balanced class distribution. After preprocessing, six traditional ML models including Logistic Regression, DTC, Random Forest, KNN, SVM and Gradient Boosting were used to predict the risk of stroke. To further increase the accuracy of the forecast. Later, ensemble models also applied to further enhance performance. Finally, ANN was applied to capture complex patterns in the data. The results shown that the ensemble model significantly improves the prediction performance compared to the individual ML models. Later a hybrid ML and DL ANN is proposed. The proposed hybrid model combining ML best model and ANN technique improved the accuracy in predicting stroke and provide a more reliable system for early detection of stroke. It is observed that the hybrid method will outperform both individual ML models and ensemble models.