Time series forecasting with high stakes: A field study of the air cargo industry

TL;DR

This field study tackles mid-term air cargo demand forecasting at the origin-destination level, addressing high volatility and revenue impact by integrating a mixture of statistical and neural models. The authors evaluate three architectures—DNN-LADD, NBEATS, and TFT—within a meta-learning framework (MAML) and apply a mixture-of-experts selection per O&D to optimize predictions over a six-month horizon. Key contributions include a practical MOE framework, meta-learning for low-data O&Ds, and validation on real carrier data with RMSE and WnRMSE metrics showing improvements over industry baselines. The work demonstrates tangible benefits for capacity allocation and revenue management, with a clear path to extending the approach to additional targets and providing attribution of revenue uplift from forecast-driven decisions.

Abstract

Time series forecasting in the air cargo industry presents unique challenges due to volatile market dynamics and the significant impact of accurate forecasts on generated revenue. This paper explores a comprehensive approach to demand forecasting at the origin-destination (O\&D) level, focusing on the development and implementation of machine learning models in decision-making for the air cargo industry. We leverage a mixture of experts framework, combining statistical and advanced deep learning models to provide reliable forecasts for cargo demand over a six-month horizon. The results demonstrate that our approach outperforms industry benchmarks, offering actionable insights for cargo capacity allocation and strategic decision-making in the air cargo industry. While this work is applied in the airline industry, the methodology is broadly applicable to any field where forecast-based decision-making in a volatile environment is crucial.

Figures (1)

• Figure 1: Illustration of look around departure date (LADD) method using the bidirectional LSTM