X-Random Forest: An Explainable AI Framework for Early Stage Diabetes Prediction with Hyperparameter Tuning and Cross-validation

Diabetes currently affects over 422 million people worldwide, presenting a significant global health challenge. While existing methods allow for the detection of diabetes through blood sugar levels, these methods often require expert analysis and fail to provide insights into the underlying causes and contributing factors. This limitation can result in missed opportunities for preventive measures, leading to new cases of diabetes. The complexity of diabetes is further compounded by the difficulty in identifying the most critical features among a limited set of factors. This study seeks to address these challenges by employing Artificial Intelligence to identify key factors and assess their impact on diabetes onset. We propose an Explainable Random Forest (XRF) model that not only predicts diabetes status with an impressive 98.59% accuracy, thanks to hyperparameter tuning and 10-fold cross-validation, but also enhances the interpretability of the model through Explainable Artificial Intelligence (XAI). By identifying and emphasizing the most important features, the study aims to guide potential patients in taking preventive measures and help those already affected to better manage their condition.