Machine Learning Approaches for Stroke Detection and SMOTE for Imbalanced Data

The majority of strokes will be caused by an unanticipated blockage of pathways by the heart and brain. To provide analytical data backing for timely, patient stroke prevention and detection, by creating a highly accurate and well-risk prediction model based on stroke risk factors using machine learning models (ML). We used numerous ML algorithms to analyse influencing factors to perform early detection, timely diagnosis, and timely treatment. In this paper, we have applied Decision Tree Classifier (DTC), K-Nearest Neighbours (KNN), and Logistic Regression (LR) for the analysis. The results were misleading because the original dataset had a data imbalance. To tackle this issue, the synthetic minority oversampling technique (SMOTE) was used in a sampling technique. There is also the use of k-fold cross-validation to extract more information about the performance of the algorithm and additionally comparing the accuracy of the three models. Therefore, this model can be used to support the decision-making of health care professionals.