Abstract
STL, which stands for Seasonal and Trend decomposition using Loess, is a technique used to decompose a time series into its underlying components: trend, seasonal, and remainder. In this study, STL has been combined with the AutoRegressive Integrated Moving Average, ARIMA model in an effort to improve the forecast performance on seasonal time series. The proposed algorithm used STL decomposition to isolate the trend, seasonal and remainder components within the time series data. ARIMA or SARIMA models were then independently fitted to each component to capture their dynamics. Finally, the component-wise forecasts were aggregated to generate the final overall forecast. Forecast performance was compared with the SARIMA model using metrics such as MAE, RMSE and MAPE. Based on a preliminary case study by using atmospheric carbon dioxide concentration data from Mauna Loa, Hawaii, the findings suggest that the proposed algorithm offers a viable alternative for improving forecast performance in seasonal data.
Reference11 articles.
1. Box G.E.P., Jenkins G.M., Time series analysis forecasting and control (Prentice-Hall, Englewood Cliffs, 1976)
2. Zhou J., Fang Q., Zhu H., Song H., Wang Y., Forecasting of the incoming dustcarts of a waste transfer station based on Sarima Model, in the 35th Chinese Control and Decision Conference, CCDC, 20-22 May 2023, Yichang, China, 4953–4958 (2023)
3. Short-term offshore wind speed forecast by seasonal ARIMA - A comparison against GRU and LSTM
4. Dagum E.B., Bianconcini S., Seasonal adjustment methods and real time trend-cycle estimation, section 1.2.2 (Springer, Switzerland, 2016)