/Triple-Exponential-Smoothing-Method-for-Temperature-Forecasting

[IEEE AP-S/USNC-URSI 2021] Propose a robust time sequence model using triple exponential smoothing to predict ground-based air temperature values via historical values.

Primary LanguagePython

Day-ahead Forecasts of Air Temperature

With the spirit of reproducible research, this repository contains all the codes required to produce the results in the manuscript:

H. Wang, M. S. Pathan, Y. H. Lee, and S. Dev, Day-ahead Forecasts of Air Temperature, Proc. IEEE AP-S Symposium on Antennas and Propagation and USNC-URSI Radio Science Meeting, 2021.

Executive summary

Climate change is a phenomenon that can affect many departments including health, development, and planning. In this reseach, we proposed a robust time sequence model using triple exponential smoothing method to predict ground-based air temperature values using historical values. Our model quantitively achieved improvement in capturing the seasonal variability of temperature and conducted quantitative evaluation on RMSE values.

Code

All codes are written in python3.

  • forecasting-example1.py: Computes the forecasting performance and plots the result for sample example 1.
  • forecasting-example2.py: Computes the forecasting performance and plots the result for sample example 2.
  • benchmarking.py: Computes the performance of our proposed method, along with other benchmarking methods.

Results

The results are stored in the folder ./results/.

  • prediction-index146.PDF: Plot of forecasting example 1.
  • prediction-index20.PDF: Plot of forecasting example 2.
  • comparison.txt: Text file that details the performance of the various benchmarking methods.

Datasets

The dataset used in our case study can be found in the folder ./data/.