시계열 분석을 활용한 지목 변화 추정 분석

Abstract

This study analyses the current status of land categories and applies cadastral statistics to a time series model to identify and predict future changes. The cadastral statistics from 2002 to 2023 were used to analyse land use status and estimate future changes. Before applying the data to the prediction model, a correlation between the number of parcels and their areas was analysed. Additionally, an ADF test was conducted, followed by two differencing steps to maintain stationarity. To determine the proper lag for the model, several lags were tested, and lag 4 was identified as the best lag based on RMSE comparison. Consequently, the data were applied to the model, and the future status of land categories was estimated. According to the comparison of estimated results and real data, it was able to distinguish the trend of land categories despite small differences. The analysis revealed that some land categories, particularly those related to infrastructure or agriculture, exhibit similar trends in future changes. This finding highlights that such time series analysis and prediction methods are useful for estimating short-term changes in land categories, which can be meaningful in designing improved land category models.