Development of a Deep Learning Model for the Early Prediction of Stroke

Abstract

This study presents the development of a deep learning model for the prediction of stroke using a healthcare dataset. The dataset was thoroughly preprocessed, encompassing feature extraction, data partitioning, and missing value management, to ensure optimal conditions for model training and evaluation. Regularisation techniques were included into a Convolutional Neural Network (CNN) architecture that was carefully designed with layers tuned for classification tasks in order to decrease overfitting and enhance generalisation. The model's performance was comprehensively assessed using a 10-fold cross-validation technique, which produced metrics such as accuracy, precision, recall, F1-score, and the Area Under the Curve (AUC). The findings showed that the CNN's average accuracy, precision, recall, and F1-score were 94.88%, 94.28%, and 95.15%, respectively. The calculated AUC of 0.96 revealed the model's high discriminating capacity. These results show the model's resilience and dependability across several data subsets, suggesting its potential for efficient stroke detection.