Cross-temporal forecast reconciliation at digital platforms with machine learning
Conference
65th ISI World Statistics Congress 2025
Format: IPS Abstract - WSC 2025
Keywords: coherent forecasting, machine learning, platform, time series
Session: IPS 925 - Machine Learning improved Time Series Analysis
Monday 6 October 10:50 a.m. - 12:30 p.m. (Europe/Amsterdam)
Abstract
Time series to be forecast are oftentimes naturally organized in a hierarchical structure. In this paper, we consider platform applications as prime examples. Platform businesses operate on a digital core, and their decision-making requires high-dimensional accurate forecast streams at different levels of cross-sectional (e.g., geographical regions) and temporal aggregation (e.g., minutes to days). It also necessitates coherent forecasts across all hierarchy levels to ensure aligned decision-making across different planning units such as pricing, product, controlling, and strategy. Given that platform data streams feature complex characteristics and interdependencies, we introduce a non-linear hierarchical forecast reconciliation method that produces cross-temporal reconciled forecasts in a direct and automated way through popular machine learning methods. The method is sufficiently fast to allow forecast-based high-frequency decision-making that platforms require. We empirically test our framework on unique, large-scale streaming datasets from a leading on-demand delivery platform in Europe and a bicycle-sharing system in New York City.