Protecting and Mitigating Against Cyber Threats

26Applications of Leveraging Diverse Machine Learning Models for Heart Stroke Prediction and its Security Aspects in Healthcare

Busa Shannu Sri¹, Kotha Dinesh Sai¹ and U. M. Gopal Krishna^2*

¹School of Computer Science & Engineering, VIT-AP University, Amaravati, India

²VIT-AP School of Business, VIT-AP University, Amaravati, India

Abstract

Accordingly, early prediction of any severe condition, for example, heart stroke, may play a vital role in improving the chances of survival and betterment of quality of life. This paper comes up with a multimodal fusion approach applied to Prediction of heart stroke from a dataset with 5110 rows and seven features, including BMI, presence of hypertension and heart disease, average glucose level, and smoking status; 7 classification models applied, including Logistic Regression, SVM, k-Nearest Neighbor, Decision Tree, Bernoulli Naive Bayes, XGBoost, and Artificial Neural Networks. This is the best-estimated multimodal fusion model with, the best accuracy, i.e., 93.74%, in integrating all algorithm predictions, which uses, for example, the power of logistic regression in identifying key feature relationships and neural networks, which learn complicated patterns. It would be robust to noise and outliers, making predictions of heart strokes with less number of features. Therefore, multimodal fusion increases the prediction accuracy tremendously and assures that the proposed tool will be very effective in saving lives.

Keywords: Data analysis, ...