The M competitions have been instrumental in advancing the theory and practice of forecasting, with the last one, the M5 competition ending on June 30 of this year. It included two separate challenges the Accuracy and the Uncertainty tracks to distinguish and emphasize their distinct significance for the forecasting field. The uncertainty challenge attracted 1,137 participants in 892 teams from 94 countries, submitting close to 10,000 forecasting entries to Kaggle that hosted the competition. A draft paper describing the uncertainty competition, the data used, the winning teams and their performance as well and its major findings can be found here. The data, benchmarks, and submissions made to the M5 forecasting competition can be found here.
Similar to the M4 competition there will be a special issue of the International Journal of Forecasting (IJF) exclusively devoted to all aspects of the M5. The special issue will include invited methodological papers, based on the top-performing submissions and contributed discussion papers. The purpose of the call is to attract high-quality discussion papers. Such discussion papers could focus on the following (indicative) areas, possibly by exploiting the available data and presenting additional analysis where applicable.
- Applicability and implications of results – Views of practitioners and academics.
- Recent advances in statistical and Machine Learning methods used for probabilistic forecasting with an emphasis on retail sales forecasting applications.
- Insights on the factors and modeling strategies that determined the performance of the winning submissions.
- Insights on the difficulties present in forecasting the tails of the uncertainty distribution, especially at low aggregation levels.
- Hierarchical forecasting and coherence in the area of probabilistic forecasting – Applicability of such approaches on large-scale retail applications.
- Insights on the importance of the exogenous and explanatory variables.
- Insights on the importance of data pre-processing, forecast adjustments, and cross-validation strategies.
We invite you to submit your interest to contribute a discussion paper for the M5 special issue (Uncertainty Track) by sending us a proposal (500 words maximum) by January 15, 2021 to Fotios Petropoulos. We will then evaluate all proposals, and notify the authors by January 31, 2021. The selected papers are to be submitted to the journal’s submission management system by April 30, 2021. The maximum length for each full paper should not exceed 5000 words. All submitted papers will follow IJF’s rigorous double-blind refereeing process.
For further information about the Special Issue, please contact the guest editors:
• Dr. Fotios Petropoulos, University of Bath, UK, [email protected]
• Professor Spyros Makridakis, University of Nicosia, Cyprus, [email protected]
• Dr. Evangelos Spiliotis, National Technical University of Athens, Greece, [email protected]