Collaboration of clustering and classification techniques for better prediction of severity of heart stroke using deep learning

Abstract

Our research aims to present a comprehensive study of machine learning algorithms and deep learning advancements in medical field systems for decision making. Present study examines the idea of extracting most important risk factors from given medical data, which has major impact in the increase of severity condition of heart stroke. Three experimental prediction models are developed when k-means clustering is collaborated with classification which includes machine learning algorithms like Naïve Bayes, Decision Tree and a deep learning algorithm Artificial Neural Network. A detailed comparison analysis is done by evaluating performance metrics like sensitivity, specificity, accuracy, and AUC-ROC scores. Out of the three, k-means with Artificial Neural Network model outperformed with sensitivity 0.89, specificity 0.89, and accuracy of 0.90 in comparison with machine learning classifiers. The challenges of perfect balancing of sensitivity and specificity is achieved by AUC-ROC score of 0.96, which is the best possible result till now.