An efficient random forest algorithm-based telemonitoring framework to predict mortality and length of stay of patients in ICU

In all phases of care, patient monitoring is essential. In particular patient monitoring in Intensive Care Units (ICUs) can lower complications and morbidity while improving the standard of care by permitting hospitals to provide better-quality, more economical patient care. The creation and verification of ICU mortality and duration of stay prediction models. The Artificial Intelligent and Vulture based Remote Patient Monitoring (AIV-RPM) framework is proposed to track patients' health status and generates timely alerts, recommendations, or reports when dangerous medical conditions are anticipated. Additionally, pre-processing is employed to remove the noise and errors using Anisotropic Diffusion Filter Based Unsharp Masking and crispening (ADF-UMC). Extract the relevant features from the dataset using Grey-Level Co-Occurrence Matrix (GLCM). Moreover, update vulture fitness function in the RF classifier for accurate prediction of LOS and mortality in ICU. The experimental outcomes of the designed model were validated with other prevailing models. Finally, the designed model gained 98.04% accuracy, 99.8% precision, 99.9% sensitivity, 97.89% F1-Score, 99.63% specificity, and 99.89% AUC for 20 data file sizes. While comparing other models designed model gained 2% greater results while comparing others models.