A Prediction of Complications in Sickle Cell Disease Patients using Regularized Random Forest

In recent years, the requirement of real-time healthcare monitoring systems has improved, particularly for personalized services. Sickle Cell Disease (SCD) is an inherited illness causing severe Red Blood Cells (RBCs) and needs continuous monitoring. Sickle Cell pain in SCD patients often leads to unplanned hospitalizations, severely impacting quality of life and overall health. Hence, this study proposes a Machine Learning (ML) based Regularized Random Forest (RRF) model for accurate prediction and classification of complications in SCD patients. This prediction process starts with collecting input images from ErythrocytesIDB dataset and the collected images are further pre-processed by using median filter, where it converts blood smear images into grayscale. Next, the features are extracted from RBCs by using Principal Component Analysis (PCA). Then, hyperparameter tuning uses Hyper Reference Engine (HRE), Grid Search (GS) and Random Search (RS) for the cross-validation of hyperparameters. Finally, RRF classifier is introduced with L1 and L2 regularization for the classification. The result analysis showed that proposed RRF model predicts SCD complications with exceptional accuracy (${9 9. 9 8 \%}$), precision (${9 9. 6 5 \%) \text {, and f1 score }}$ ${(9 9. 2 0 \%)}$ respectively when compared with existing Convolutional Neural Network.