함수형 시계열 자료를 위한 추정 및 예측모형과 응용

Abstract

Carbon dioxide (CO2) in the atmosphere is a greenhouse gas that absorbs radiant energy emitted from the earth's surface and maintains a suitable temperature for life on the earth. However, human activities such as rapid industrialization have recently increased the concentration of CO2, which has become a cause of global warming. In order to understand and monitor changes in CO2 level, many studies are learning through numerical models related to current and past CO2 level and predicting future CO2 level. In this study, we consider the CO2 data measured by the Korea Meteorological Administration as a functional time-series data model, incorporate the functional data method with the time-series method. Existing studies have used traditional time-series and longitudinal data approaches after fixing seasonal and spatial criterion to entire CO2 data. However, assuming that the CO2 level is a continuous function over the time space looks more reasonable than assuming the CO2 level is a scalar at individual time point. Therefore, a part from the traditional methods, we assume the observed data as functional data and use a functional data method to infer the dependent relationship between the functions and apply such a approach to the CO2 data collected from 2001 to 2021 in Korea. Also, we introduce forecasting method that can predict the CO2 level for a specific future time.