search for




 

Population forecasting by small region in Korea using LSTM
Journal of the Korean Data & Information Science Society 2025;36:179-89
Published online January 31, 2025;  https://doi.org/10.7465/jkdi.2025.36.1.179
© 2025 Korean Data and Information Science Society.

Da Hye Lee1 · In Hong Chang2 · Kwang Yoon Song3

123Department of Computer Science and Statistics, Chosun University
Correspondence to: This study was supported by research fund from Chosun University, 2024.
1 Researcher, Department of Computer Science and Statistics, Chosun University, Gwangju 61452, Korea.
2 Professor, Department of Computer Science and Statistics, Chosun University, Gwangju 61452, Korea.
3 Corresponding author: Assistant Professor, Department of Computer Science and Statistics, Chosun University, Gwangju 61452, Korea. E-mail: csssig@chosun.ac.kr
Received November 20, 2024; Revised December 13, 2024; Accepted December 13, 2024.
This is an Open Access article distributed under the terms of the Creative Commons Attribution Non-Commercial License (http://creativecommons.org/licenses/by-nc/3.0) which permits unrestricted non-commercial use, distribution, and reproduction in any medium, provided the original work is properly cited.
Abstract
South Korea is experiencing an unprecedented crisis of regional population decline. The fundamental causes of depopulation include a natural decline in birth rates and aging. However, various complex factors such as a lack of employment opportunities, insufficient cultural and educational environments, among others, contribute to the crisis. While research on depopulation has focused on forecasting birth rates, the number of women of childbearing age, and marriage rates, as well as developing depopulation indices considering multiple factors, these studies have mainly been conducted on a large scale, focusing on large cities. This study aims to propose a short-term population forecast model using monthly population data at the level of large cities (metropolitan cities, do, and special self-governing provinces) and small cities (si, gun, and gu). The study discusses time series analysis for short-term population forecast, comparing the autoregressive integrated moving average (ARIMA) and the long short-term memory (LSTM). Both ARIMA and LSTM models are fitted with min-max normalization applied to the population data to enhance accuracy. The LSTM structure considers a model with two stacked LSTMs combined with hidden layer. ARIMA fits well on a regional basis, while LSTM shows reasonable performance except for some cities. Although this study is limited to population forecast models, it is suggested that future research could expand to incorporate information on budgets for local governments and regional infrastructure. By combining such data, the study could extend to model research aimed at improving the system for regional affection, ultimately contributing to the enhancement of the current ‘Hometown Love e-um’.
Keywords : ARIMA, LSTM, population forecast, small region