Performance Comparison of Machine Learning Algorithms Using Oversampling Methods to Predict Childhood Anemia

Authors : Kadriye Filiz Balbal

Pages : 1-11

Doi:10.7212/karaelmasfen.1555212

View : 322 | Download : 327

Publication Date : 2025-11-19

Article Type : Research Paper

Abstract :Childhood anemia is a major health problem. Anemia, which is common in preschool-aged children, causes physical and mental developmental delays in this age group. It is important to address and investigate these preventable and treatable health problems using state-of-the-art methods such as artificial intelligence. Therefore, this study employs machine learning techniques, a subfield of artificial intelligence, to predict the anemia levels in children aged 0–59 months in Nigeria. To address the issue of data imbalance, which can cause problems in estimating childhood anemia levels, the SMOTE and ADASYN oversampling techniques were employed in this study. In the analyses performed with the newly obtained balanced data, it was observed that the SMOTE and ADASYN methods performed significantly better than the results obtained with imbalanced data for all ML models. When the average results of all ML algorithms used in this study in terms of accuracy, precision, recall, and F1 score metrics are compared to the oversampling methods, the most successful result in terms of all metrics was obtained with the SMOTE method.
Keywords : Yapay zeka, makine öğrenmesi, ADASYN, SMOTE, çocukluk çağı anemisi

ORIGINAL ARTICLE URL