Early health prediction framework using XGBoost ensemble algorithm in intelligent environment

Amidst the COVID-19 humanitarian catastrophe, the Internet of Things and Artificial Intelligence (AI) are premier technologies in the healthcare domain that have emerged to a great extent. This global health emergency highlights the need to bolster current healthcare systems for future preparedness. Conspicuously, the current paper presents a non-invasive AI-empowered model for passive health monitoring and predicting viral C-19 infection in the home environment. It consists of four notable layers: fully automated data acquisition, data analysis and Bayesian probabilistic classification, temporal COVID-19 severity prediction, and communication layer. These layers include IoT sensors embedded in the intelligent toilet system to collect required data, processes and analyses of the urine parametric data at the fog layer, and forecasting the COVID-19 severity using the XGBoost machine learning model at the cloud layer. The model has been evaluated over 53,550 data instances in a simulated environment for implementation purposes. The results implied that the proposed AI framework outperformed state-of-the-art strategies in terms of temporal approximation (94.53 s), reliability (92.69%), stability (0.89%), and predictive performance analysis (95.26%) metrics.