Forking-Sequences

2025-10-06

AI Summary

In time series forecasting, models exhibit significant output fluctuations across different forecast creation dates (FCDs), undermining prediction reliability and downstream decision-making. To address this, we propose the first systematic “forked sequence” framework: it jointly encodes and decodes full prediction sequences from multiple FCDs, thereby unifying temporal dependency modeling with cross-FCD stability learning. By sharing parameters across FCDs, the method enables smoother gradient updates, inherently supports ensemble-based variance reduction, and substantially improves inference efficiency. We validate it across diverse architectures—including MLP, RNN, LSTM, CNN, and Transformer—using time-series-aware cross-validation. Experiments span 16 benchmark datasets (M1–Tourism), demonstrating average improvements of 28.8%–37.9% in prediction stability across FCDs; notably, Transformer variants achieve an average 8.8% gain. This work establishes the first scalable, architecture-agnostic modeling paradigm for enhancing temporal forecasting stability.

Technology Category

• Machine Learning: Time-Series/Data Streams
• Planning, Routing, and Scheduling: Temporal Planning
• Reasoning under Uncertainty: Sequential Decision Making

Application Category

• User Modeling, Personalization and Recommendation: Federated recommendation systems and personalization
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

While accuracy is a critical requirement for time series forecasting models, an equally important (yet often overlooked) desideratum is forecast stability across forecast creation dates (FCDs). Even highly accurate models can produce erratic revisions between FCDs, undermining stakeholder trust and disrupting downstream decision-making. To improve forecast stability, models like MQCNN, MQT, and SPADE employ a little-known but highly effective technique: forking-sequences. Unlike standard statistical and neural forecasting methods that treat each FCD independently, the forking-sequences method jointly encodes and decodes the entire time series across all FCDs, in a way mirroring time series cross-validation. Since forking sequences remains largely unknown in the broader neural forecasting community, in this work, we formalize the forking-sequences approach, and we make a case for its broader adoption. We demonstrate three key benefits of forking-sequences: (i) more stable and consistent gradient updates during training; (ii) reduced forecast variance through ensembling; and (iii) improved inference computational efficiency. We validate forking-sequences' benefits using 16 datasets from the M1, M3, M4, and Tourism competitions, showing improvements in forecast percentage change stability of 28.8%, 28.8%, 37.9%, and 31.3%, and 8.8%, on average, for MLP, RNN, LSTM, CNN, and Transformer-based architectures, respectively.

Problem

• Improving forecast stability across different creation dates
• Reducing erratic revisions between forecast creation dates
• Enhancing training stability and computational efficiency

Innovation

• Jointly encodes and decodes entire time series
• Improves forecast stability across creation dates
• Enhances training and inference computational efficiency