|
Exploring Window Size Effects in Window-sliding ARIMA Models for Future Use in PM2.5 Data Imputation |
|---|---|
| รหัสดีโอไอ | |
| Creator | Jularat Chumnaul |
| Title | Exploring Window Size Effects in Window-sliding ARIMA Models for Future Use in PM2.5 Data Imputation |
| Contributor | Jularat Chumnaul, Kasikrit Damkliang |
| Publisher | Thai Society of Higher Education Institutes on Environment |
| Publication Year | 2569 |
| Journal Title | EnvironmentAsia |
| Journal Vol. | 19 |
| Journal No. | 1 |
| Page no. | 1-17 |
| Keyword | Holt-Winters, Box-Jenkins, Auto-regressive integrated moving average, Window-sliding, Imputation |
| URL Website | http://www.tshe.org/ea/index.html |
| Website title | EnvironmentAsia |
| ISSN | 1906-1714 |
| Abstract | Accurately predicting PM2.5 concentrations is crucial for effective strategic planningto reduce PM2.5 pollution. Precise PM2.5 forecasts offer significant benefits for managingpollution levels. This study examines the forecasting performance of window-sliding ARIMA(WS-ARIMA) models with varying historical window sizes, using complete PM2.5 datasetsas a basis for evaluating their potential application in imputing long-period missing data. Astandard ARIMA (1,1,2) model is compared to WS-ARIMA models using window sizes from7 to 365 days across three forecast horizons: 7, 14, and 31 days. Evaluation metrics, includingRMSE, MAE, and MAPE, demonstrate that very short and very long window sizes yield inferiorperformance due to either overfitting noise or oversmoothing important temporal patterns. Incontrast, WS-ARIMA models with window sizes of 150 and 180 days consistently outperformother configurations, providing optimal predictive accuracy across all forecast periods. Atwo-way ANOVA confirms that both window size and time horizon have a significant effecton model performance; these findings indicate that WS-ARIMA models, when equipped withappropriately selected window sizes, are particularly effective for imputing long-term gaps inPM2.5 data. Therefore, the results of this study provide practical guidance for environmentalresearchers who are looking for reliable statistical tools for restoring missing air quality data. |
