Most recent data guides better long term time series predictions
Most-Recent Anchoring with Recurrent Ordering for Time Series Forecasting
Machine Learning
Summary
Forecasting future values from time series data is hard because recent information often matters more than older data. The authors propose a model called MARO that looks at the newest information first and then gradually incorporates older data, keeping the focus on the latest trends. This approach uses the same processing step repeatedly without extra parameters, efficiently emphasizing recent data while still considering the past. Experiments show that this method improves forecasting accuracy over different time spans and datasets.
What this means in practice
- •For financial analysts: Produce more accurate long-term market trend forecasts by focusing model computations on the most recent price movements.
- •For energy demand planners: Improve short- and long-term electricity demand predictions by weighting the latest consumption data more strongly in the forecasting model.
Authors
Jung Min Choi, Ngoc Son Le, Ibram Abdelmalak, Vijaya Krishna Yalavarthi, Lars Schmidt-Thieme
Abstract
Long-term forecasting models commonly process all patches in a look-back window using the same fixed stack. Older contextual patches and recent evidence therefore receive the same computational depth. Yet the information closest to the forecast and the more distant context do not contribute equally. Uniform processing leaves this distinction unexpressed in the architecture. We propose MARO, a Most-Recent Anchoring with Recurrent Ordering model that processes the look-back window from the most recent patch to the oldest. The most recent patch serves as the anchor. It initializes the latent state and conditions each subsequent step, so older patches are folded into a representation that remains centered on recent evidence. A single shared module is reused at every step, so extending the scan further into the past introduces no additional parameters. Intermediate states retained during the scan allow the forecast head to weigh short and long portions of the history separately. This expresses recency through the order of recurrent refinement. Extensive experiments across multiple real-world time series datasets show that MARO achieves state-of-the-art performance on both long-term and short-term forecasting tasks.Ablation studies examine the contribution of the main architectural components.