The wisdom of crowds in forecasting at high-frequency for multiple time horizons: A case study of the Brazilian retail sales
Main Article Content
Abstract
This case study compares the forecasting accuracy obtained for four daily Brazilian retail sales indices across four prediction horizons. The performance of traditional time series forecasting models, artificial neural network architectures and machine learning algorithms were compared in order to evaluate the existence of a single best-performing model. Subsequently, ensemble methods were included in the model comparison to determine if accuracy could be improved. Evidence from this case study suggests that a consistent forecasting strategy exists for the Brazilian retail indices by applying both seasonality treatment for holidays and calendar effects and by using an ensemble method whose main inputs are the predictions of all models with calendar variables. This strategy was consistent across all 16 index and time horizon combinations, as ensemble methods either outperformed the best single models or showed no statistical difference from them in a Diebold-Mariano test.
Article Details

This work is licensed under a Creative Commons Attribution 4.0 International License.
Articles published from Vol. 22 No. 4 (October–December 2024) onward are licensed under the Creative Commons Attribution 4.0 International License (CC BY 4.0). Earlier articles carry the license stated in their PDF.