arXiv:2607.01959v1 Announce Type: cross Abstract: We propose a model agnostic methodology to measure lag relevance in machine learning forecasting models applied to univariate time series. Particularly, we are working in the context of time series using the frameworks of Ghost variables and Shapley values, together with additive importance measures, to introduce the auto-relevance and partial auto-relevance functions as the lag importance values. Additionally, we propose a novel method to replace absent features in coalition based methods with a one step forecast from the same model. We evalua
The proliferation of machine learning in time series forecasting necessitates more robust methods for understanding feature contributions and improving model interpretability.
This research provides a foundational tool for enhancing the reliability and transparency of AI-driven forecasting, critical for integrating these models into sensitive applications.
The introduction of autorelevance functions and improved methods for handling absent features could lead to more accurate and trustworthy time series models.
- · Machine Learning Researchers
- · Financial Forecasting Firms
- · Supply Chain Management
- · Time Series Data Analysts
- · Black-box AI Models
- · Manual Feature Engineering
Improved interpretability of time series forecasting models by identifying key lag contributions.
Increased adoption of complex machine learning models in industries requiring high explainability due to enhanced transparency.
Further development of interpretable AI frameworks, potentially standardizing how 'ghost variables' and 'Shapley values' are applied in novel contexts.
This signal links to a primary source. Continuum Brief monitors and indexes it as part of the live intelligence stream — we do not republish source content.
Read at arXiv cs.LG