From Distributional to Quantile Neural Basis Models: the case of Electricity Price Forecasting

2025-09-17

AI Summary

Addressing the challenge of balancing model interpretability and predictive accuracy in multi-horizon probabilistic electricity price forecasting, this paper proposes the Quantile Neural Basis Model (Q-NBM). Q-NBM embeds the structural interpretability of quantile generalized additive models into an end-to-end neural network framework by employing shared basis function decomposition and weight factorization, thereby explicitly modeling nonlinear feature contributions to each output quantile. Unlike conventional approaches, it imposes no prior assumptions on the error distribution and supports both multi-step quantile regression and joint location-scale-shape modeling. Evaluated on day-ahead electricity price forecasting, Q-NBM achieves predictive performance comparable to state-of-the-art distributional and quantile neural networks, while delivering both global and local interpretability of feature effects—effectively bridging the gap between deep learning and interpretable probabilistic forecasting.

Technology Category

• Machine Learning: Calibration & Uncertainty Quantification
• Reasoning under Uncertainty: Relational Probabilistic Models
• Natural Language Processing: Interpretability, Analysis, and Evaluation of NLP Models

Application Category

• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• User Modeling, Personalization and Recommendation: Explainable and interpretable methods for personalization
• Web Mining and Content Analysis: Large pretrained models with web data

Abstract

While neural networks are achieving high predictive accuracy in multi-horizon probabilistic forecasting, understanding the underlying mechanisms that lead to feature-conditioned outputs remains a significant challenge for forecasters. In this work, we take a further step toward addressing this critical issue by introducing the Quantile Neural Basis Model, which incorporates the interpretability principles of Quantile Generalized Additive Models into an end-to-end neural network training framework. To this end, we leverage shared basis decomposition and weight factorization, complementing Neural Models for Location, Scale, and Shape by avoiding any parametric distributional assumptions. We validate our approach on day-ahead electricity price forecasting, achieving predictive performance comparable to distributional and quantile regression neural networks, while offering valuable insights into model behavior through the learned nonlinear mappings from input features to output predictions across the horizon.

Problem

• Interpret neural network mechanisms in probabilistic forecasting
• Introduce interpretable Quantile Neural Basis Model
• Validate model on electricity price forecasting performance

Innovation

• Quantile Neural Basis Model with interpretability
• Shared basis decomposition and weight factorization
• Nonparametric approach avoiding distributional assumptions