Causal feature selection framework for stable soft sensor modeling based on time-delayed cross mapping

2026-01-20

Advanced Engineering Informatics

AI Summary

This study addresses the limitations of existing causal feature selection methods, which often neglect time delays and dynamic dependencies among variables in industrial processes, leading to insufficient accuracy and robustness in soft sensor models. To overcome this, we propose a delay-aware causal feature selection framework based on time-delayed cross-mapping. By reconstructing state spaces to capture lagged interdependencies, the framework introduces Time-Delayed Convergent Cross-Mapping (TDCCM) and Time-Delayed Partial Cross-Mapping (TDPCM) for inferring overall and direct causal relationships, respectively. An adaptive strategy is further developed to automatically determine causal thresholds based on validation set performance, enabling data-driven feature selection. Evaluated on two real-world industrial datasets, the proposed approach significantly improves prediction accuracy, with TDCC

Technology Category

• Machine Learning: Causal Learning
• Reasoning under Uncertainty: Causality
• Knowledge Representation and Reasoning: Action, Change, and Causality

Application Category

• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs
• Web Mining and Content Analysis: Models for Web evolution
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications

Problem

• causal feature selection
• soft sensor modeling
• time delay
• interdependent variables
• causal inference

Innovation

• time-delayed cross mapping
• causal feature selection
• soft sensor modeling
• convergent cross mapping
• partial cross mapping