Forecasting high levels of PMnullnull in Korea based on the principal expectile component regression

Abstract

As the level of fine dust has risen sharply recently, many studies has been conducted to analyze the data. Since exposure to fine dust is related to the occurrence of cardiovascular diseases and respiratory, it can make the mortality rate increase. Therefore, it is important to predict the extreme level of fine dust. In this paper, we consider a regression model based on the principal expectile analysis. Compare to the conventional principal component analysis, principal expectile analysis can capture variations around the tail of the data. By so doing, we predict 'Bad' cases of the PM10 level of 25 districts in Seoul, South Korea and compare the results with the classical principal component regression. From the results, we observe that the proposed model predicts the extreme level of fine dust better than the existing model.