Optimal forecast reconciliation with time series selection

Code for paper titled "Optimal forecast reconciliation with time series selection".

Introduction

• R: R functions used for performing the proposed methods.
• Python: Python functions used for solving optimization problems via Gurobi.
• code: Code files for reproducing the results presented in the paper.
  • simulation.R: generates simulation data for two setups.
  • labour.R: organizes the unemployment data in Australian labour force application.
  • tourism.R: organizes the tourism flow data in Australian domestic tourism application.
  • baseforecast.R: generates base forecasts for both simulation datasets and empirical datasets.
  • subset_reconciliation.R, intuitive_reconciliation.R, and lasso_reconciliation.R: perform proposed reconciliation methods with time series selection as well as benchmark reconciliation methods.
  • evaluation.R: evaluates forecast performance.
  • results.R: generates results presented in the paper.
• data: Two datasets used in applications: Australian labour force, and Australian domestic tourism.
• paper: Source files for manuscript.
  • The paper is written using Quarto. Tex file and PDF file are generated by rendering the hf_selection.qmd file.