A Data-Driven Approach to Detect Dehydration in Afghan Children Using Deep Learning

Child dehydration is a significant health concern, especially among children under 5 years of age, as they are more susceptible to conditions such as diarrhea and vomiting. In Afghanistan, the impact of severe diarrhea on child mortality is exacerbated by dehydration. However, there is a notable gap in the research landscape, particularly a lack of exploration into the potential of deep learning techniques for diagnosing dehydration in Afghan children under five years of age. To address this gap, our study leveraged three powerful classifiers: Deep Learning (DL), eXtreme Gradient Boosting Classifier (XGBoost), and K-Nearest Neighbors (KNN). We developed a predictive model using a comprehensive dataset of sick children obtained from the Afghanistan Demographic and Health Survey (ADHS). The primary objective of our research was to accurately determine the dehydration status of children under 5 years of age, providing crucial insights for healthcare professionals. Among all the classifiers we evaluated, the DL approach emerged as the most effective, achieving a remarkable accuracy of 99% on both the test and validation sets, along with an impressive Area under the Curve (AUC) score of 0.99. The KNN classifier also performed solidly, with a consistent 90% accuracy across all evaluation metrics. The XGBoost classifier demonstrated a remarkable precision rate of 98%, highlighting its robustness. Our DL model has the potential to significantly assist healthcare professionals in promptly and accurately identifying dehydration in children under five, leading to timely interventions that can substantially reduce the risk of severe health complications. This study showcases the promising application of deep learning approach in improving the early diagnosis of dehydration specifically in the context of Afghan children, contributing to enhanced healthcare outcomes and saving children life.