Predictive analysis of birth intention using machine learning

Abstract

This study aimed to identify key determinants of birth intention and construct a predictive model to classify individuals at risk using machine learning. In this study, we analyzed data using machine learning techniques to identify significant predictors of birth intention and to develop a predictive model for future birth intention. Data from 2580 subjects were analyzed using the Korean Early Childhood Education & Care Panel in January 2025, employing Python version 3.12.8 via Colab (google.colaboratory application). Decision tree, random forest classifier, and logistic regression models were evaluated based on precision, accuracy, recall, F1-score, and area under the curve (AUC). The Random Forest model was selected, achieving an AUC of 82%. Among the 17 features, marriage period, age, stress, prenatal weight, number of children, maternal rearing behavior, conflict with partner, marriage satisfaction, depression, and type of birth emerged as important predictors of birth intention. Maternal stress, rearing behavior, conflict with partner, and satisfaction levels were strong predictors of future birth intention. Maternal psychological status and relational factors should be addressed with appropriate social support during the postpartum period through family-centered care.