A Lightweight Multi-View Approach to Short-Term Load Forecasting

2026-02-09

AI Summary

This work proposes a lightweight multi-view approach to address the challenges of overfitting, poor stability, and sensitivity to outdated data commonly encountered in short-term load forecasting with complex models. By integrating singleton embedding with scaled temporal input ranges, the method efficiently captures temporal dependencies while introducing an innovative embedding dropout mechanism to reduce feature co-dependence, thereby enhancing both model robustness and interpretability. Extensive experiments across multiple datasets demonstrate that the proposed approach achieves competitive forecasting accuracy with significantly fewer parameters, exhibits strong resilience to noise and data sparsity, and effectively quantifies the contribution of individual features to the prediction outcomes.

Technology Category

Machine Learning: Multi-instance/Multi-view LearningComputer Vision: Multi-modal VisionPlanning, Routing, and Scheduling: Temporal Planning

Application Category

Graph Algorithms and Modeling for the Web: Graph embeddings and representation learning for Web-related graphsSemantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMsWeb Mining and Content Analysis: Large pretrained models with web data

Abstract

Time series forecasting is a critical task across domains such as energy, finance, and meteorology, where accurate predictions enable informed decision-making. While transformer-based and large-parameter models have recently achieved state-of-the-art results, their complexity can lead to overfitting and unstable forecasts, especially when older data points become less relevant. In this paper, we propose a lightweight multi-view approach to short-term load forecasting that leverages single-value embeddings and a scaled time-range input to capture temporally relevant features efficiently. We introduce an embedding dropout mechanism to prevent over-reliance on specific features and enhance interpretability. Our method achieves competitive performance with significantly fewer parameters, demonstrating robustness across multiple datasets, including scenarios with noisy or sparse data, and provides insights into the contributions of individual features to the forecast.

Problem

• short-term load forecasting
• overfitting
• model complexity
• time series forecasting
• forecast stability

Innovation

• lightweight multi-view
• single-value embeddings
• scaled time-range input
• embedding dropout
• short-term load forecasting