Enhanced Random Forest with Optimized Grey Wolf Algorithm for Missing Data Management in Internet of Medical Things

Missing data management is a critical challenge in Internet of Medical Things (IoMT) environments. Defects in data integrity and accuracy can distort the results of Artificial Intelligence (AI)-based model analyses for patient diagnosis and treatment, potentially leading to irreversible consequences in healthcare. There are various approaches to managing missing data, each with advantages and disadvantages. Random Forest (RF) is a Machine Learning (ML) model used to predict and impute missing values. It provides more accurate predictions by combining multiple decision trees simultaneously. However, this method is time-consuming, requires high and substantial computational resources, and is highly complex due to the need for hyperparameter tuning. Hyperparameter tuning in ML models requires a high level of knowledge and experience. This chapter presents a solution based on the Enhanced Gray Wolf Optimization (EGWO) algorithm for hyperparameter fine-tuning of the RF model, aiming to increase accuracy in missing data management. The WUSTL EHMS-2020 dataset is used to evaluate this method. The dataset includes IoMT environment data with both numerical and categorical missing values. The results showed that the proposed method demonstrates high performance compared to other hyperparameter optimization methods, including Random Search, Genetic Algorithm (GA), Hyperopt, Particle Swarm Optimization (PSO), Grid Search, Optuna, and the standard GWO.