Development of a Low Birth Rate Prediction Model Using Time-series and Machine Learning

(Purpose) The main goal of this study is to build a time-series prediction model to address Korea's low birth rate using machine learning and ARIMA models. It aims to predict future fertility trends and analyze socio-economic factors influencing birth rates with a focus on the MZ generation whose views on marriage and childbirth differ from previous generations. (Design/methodology/approach) This study utilizes machine learning models, including Linear Regression, Support Vector Machine (SVM), Random Forest, AdaBoost, Neural Networks, and the ARIMA time-series model. These models are trained on socio-economic data, analyzing variables like marriage rates, employment rates, and economic growth. The predictive accuracy of the models is evaluated using performance metrics such as Mean Squared Error (MSE), Root Mean Squared Error (RMSE), Mean Absolute Error (MAE), and R² to forecast future birth rates. (Findings) The findings reveal that Neural Networks exhibited the highest predictive accuracy for future birth rates, with an R² of 0.978, followed by Linear Regression and SVM. The ARIMA model effectively forecasted long-term birth rate trends, demonstrating its suitability for analyzing time-series data. Socio-economic factors such as childcare subsidies, household income, and age at first marriage were found to significantly impact the fertility rate. (Research implications or Originality) This research combines machine learning models and time-series analysis to predict Korea's low birth rate. It is to highlight the need for policies that address both economic support and societal attitudes toward marriage and childbirth. It provides a framework for policymakers to design effective interventions and emphasizes the importance of non-linear relationships between socio-economic factors and birth rates, offering a novel approach in demographic studies.