Development of a Machine Learning-Based Low Birth Weight Risk Prediction System
61st International Scientific Conference on Information, Communication and Energy Systems and Technologies, ICEST 2026, Nis, Serbia, 1 - 03 July 2026, pp.216-217, (Full Text)
- Publication Type: Conference Paper / Full Text
- Doi Number: 10.1109/icest71230.2026.11623508
- City: Nis
- Country: Serbia
- Page Numbers: pp.216-217
- Keywords: clinical decision, Low birth weight, machine learning, risk prediction
- Dokuz Eylül University Affiliated: Yes
Abstract
This study aims to predict the risk of low birth weight (LBW) using machine learning methods based on clinical, demographic, and socioeconomic data of the mother during pregnancy. The dataset was utilized, and the target variable was transformed into a binary classification problem as 'Low' (risky) and 'Normal'. In the data preprocessing phase, missing values were imputed, numerical data were scaled, and categorical data were encoded as binary. Logistic Regression, Decision Trees, Random Forest, SVC, Gradient Boosting, and XGBoost models were compared. The Gradient Boosting model was selected due to its success in capturing non-linear relationships. Considering the importance of risk detection in healthcare, a threshold value of 0.35 was used instead of the default 0.5 to create a more sensitive decision mechanism.