/NTS_forecasting_via_SGWT

Code for the paper 'Network time series forecasting using spectral graph wavelet transform'

Primary LanguageRMIT LicenseMIT

Network Time Series Forecasting using Spectral Graph Wavelet Transform

This repo contains the source code for network time series forecasting using SGWT.

For the detail, please refer to our paper: https://doi.org/10.1016/j.ijforecast.2023.08.006.

Description

  • Code

    • method.R is a code for functions used for forecasting.
    • seoulmetro.R and covidseoul.R are codes for real data analysis.

  • Data

    • seoulmetro contains data of daily number of people getting on and off the Seoul Metropolitan Subway in South Korea.
    • covidseoul contains data of daily number of newly confirmed cases in the Seoul Metropolitan Area.

Code overview

We apply various network forecasting methods to analyze real-world data, Seoul Metropolitan Subway data and Seoul Metropolitan COVID-19 data. The methods used in the analysis are

  • Nodewise ARIMA
  • LOCAAT-based method
  • GNARI
  • GNARI + LOCAAT
  • GFT-based method
  • SGWT-based method