Heart Disease prediction with Feature Sensitized Interpretable Framework for Internet of Medical Sensors

Nallakaruppan Kailasanathan¹Gangadevi Ezhilarasan²Shitharth Selvarajan^3,4*Rajesh Kumar Dhanaraj⁵Dragan Pamucar^6*Nathan Shankar⁷

• ¹Balaji Institute of Modern Management (BIMM), Pune, Maharashtra, India
• ²Loyola College, Chennai, Chennai, Tamil Nadu, India
• ³School of Built Environment, Engineering & Computing, Leeds Beckett University, Leeds, United Kingdom
• ⁴Leeds Beckett University, Leeds, England, United Kingdom
• ⁵Symbiosis International University, Pune, Maharashtra, India
• ⁶Széchenyi István University, Gyor, Gyor-Moson-Sopron, Hungary
• ⁷The University of Manchester, Manchester, England, United Kingdom

The central control system of the human body is the heart. The terms pertaining as the heart of the city or the heart of the computer is understood, for the solidity 1 Nallakaruppan et al. Running Title of heart as a part of human conversations. Challenging situations like cholesterol, high blood pressure, stress, obesity, lifestyle, smoking, and drinking challenge the functioning of the heart. The drastic increase in younger people dying of heart failure is the most undesirable and common situation in recent times. Postpandemic, this situation has become common among the age group of 25-40 years, due to the impact of medication and the post pandemic health practices. This enhances the need of the automatic interpretable framework, with feature analysis for the heart disease prediction. The proposed work provides a solution for this early heart failure situation by constantly monitoring the body's vitals through wearable and other Internet of Medical Things (IoMT) sensors and then integrating the data into the cloud. The next step entails extracting the data sets from the cloud to perform classification with the help of machine learning models such as SVM, Random Forest, Decision Tree, and Logistic Regression. The proposed work selects the Random Forest Classification model with the highest accuracy and F1-score compared to all other models with the value of 0.955, with k fold validation. Subsequently, it applies for explainability using LIME in local surrogates and SHAPELY in global surrogates. SHAPELY's decision plot describes classification behavior with the decision and other plots.