Auto-Configured Networks for Multi-Scale Multi-Output Time-Series Forecasting
2026-04-08 • Machine Learning
Machine LearningNeural and Evolutionary Computing
AI summaryⓘ
The authors address the challenge of forecasting multiple outputs from different data sources that are not perfectly aligned in time. They create a new type of neural network that looks at both short-term changes and long-term trends to improve predictions. They also develop an automated method to find the best combination of data processing, model design, and training settings while balancing prediction accuracy and model simplicity. Their approach finds a set of models that trade off error and complexity well, and tests show it works better than existing methods on both synthetic and real-world data. This helps users choose models that fit their specific needs and computing limits.
multi-source signalsmulti-output regressionneural networksmulti-scale convolutionmodel complexityPareto optimizationevolutionary algorithmhyperparameter tuningtime series forecastingmodel deployment
Authors
Yumeng Zha, Shengxiang Yang, Xianpeng Wang
Abstract
Industrial forecasting often involves multi-source asynchronous signals and multi-output targets, while deployment requires explicit trade-offs between prediction error and model complexity. Current practices typically fix alignment strategies or network designs, making it difficult to systematically co-design preprocessing, architecture, and hyperparameters in budget-limited training-based evaluations. To address this issue, we propose an auto-configuration framework that outputs a deployable Pareto set of forecasting models balancing error and complexity. At the model level, a Multi-Scale Bi-Branch Convolutional Neural Network (MS--BCNN) is developed, where short- and long-kernel branches capture local fluctuations and long-term trends, respectively, for multi-output regression. At the search level, we unify alignment operators, architectural choices, and training hyperparameters into a hierarchical-conditional mixed configuration space, and apply Player-based Hybrid Multi-Objective Evolutionary Algorithm (PHMOEA) to approximate the error--complexity Pareto frontier within a limited computational budget. Experiments on hierarchical synthetic benchmarks and a real-world sintering dataset demonstrate that our framework outperforms competitive baselines under the same budget and offers flexible deployment choices.