Dynamic Feature Extraction and Interactive Visualization: A Human-Collaborative Intelligent Analysis System for Complex Textile Manufacturing Datasets

Abstract

As the era of big data continues to evolve, the proliferation of massive, multi-dimensional datasets—particularly those generated by high-speed fiber manufacturing and intelligent textile monitoring systems—poses significant challenges to traditional data analysis methodologies. In modern intelligent sensing environments, electromagnetic signal acquisition and wireless monitoring technologies further increase the complexity and dynamic characteristics of industrial data streams, requiring more adaptive analytical frameworks. This paper introduces a novel approach that integrates machine learning-based feature extraction with interactive dynamic visualization to establish a tightly coupled human-collaborative intelligent analysis system. The proposed framework transforms conventional static feature extraction into a dynamic iterative process encompassing feature perception, model optimization, insight verification, dynamic feature engineering, importance-shift-based visualization mapping, and real-time human–machine interaction feedback. Application examples involving complex textile manufacturing datasets demonstrate that the framework not only improves the efficiency and interpretability of feature extraction for fiber quality parameters but also enhances analysts’ understanding of model behavior and decision-making logic. The results indicate broad application prospects for intelligent industrial monitoring and adaptive data analysis while providing useful references for electromagnetic sensing-assisted information processing and real-time decision support in complex manufacturing environments.