Scholarlink

<p>Artificial intelligence (AI) has emerged as a key driver in product design instruction, enabling adaptive learning experiences, personalized instructional support, and data-driven performance analysis. Traditional intelligent tutoring systems struggle to capture complex cognitive patterns and temporal dependencies within learner interaction data, limiting instructional adaptability and prediction accuracy. Deep learning approaches rely on static feature extraction and fixed optimization, leading to poor generalization when handling multimodal learning behavior. Further, the absence of dynamic tuning mechanisms restricts model responsiveness in evolving learner performance. To address these challenges, the proposed model, dynamic electric Eel multi-kernel convolutional bi-memory network (DEE-MKC-BiMN), an adaptive hybrid neural framework, enhances feedback accuracy, learning adaptability, and instructional intelligence in product design education. Outlier detection identifies and removes anomalous interaction records to preserve meaningful learning patterns, while Min–Max normalization standardizes heterogeneous features for stable model training. Data augmentation is incorporated to enhance dataset diversity and reduce overfitting by generating representative variations of learner interaction patterns. Linear discriminant analysis refines key indicators to maximize class separability. Multi-kernel convolutional (MKC) layers extract diverse spatial behavior features, and a bi-directional memory module retains sequential interaction dependencies crucial for learning progression. DEE tunes network parameters to avoid premature convergence and improve optimization stability. Evaluated on learner interaction data and activity logs from AI-supported product design environments, DEE-MKC-BiMN achieves 96.86% of accuracy, demonstrating enhanced predictive performance, faster convergence, and improved adaptability compared with conventional hybrid models. DEE-MKC-BiMN strengthens intelligent tutoring effectiveness through multi-scale feature learning and fosters superior adaptive learning outcomes.</p> Graphical abstract <p></p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Construction and effectiveness evaluation of an AI-based intelligent tutoring system for product design instruction

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年7月19日

查看全文
  • Min Shi,
  • Xiaofang Ma

摘要

Artificial intelligence (AI) has emerged as a key driver in product design instruction, enabling adaptive learning experiences, personalized instructional support, and data-driven performance analysis. Traditional intelligent tutoring systems struggle to capture complex cognitive patterns and temporal dependencies within learner interaction data, limiting instructional adaptability and prediction accuracy. Deep learning approaches rely on static feature extraction and fixed optimization, leading to poor generalization when handling multimodal learning behavior. Further, the absence of dynamic tuning mechanisms restricts model responsiveness in evolving learner performance. To address these challenges, the proposed model, dynamic electric Eel multi-kernel convolutional bi-memory network (DEE-MKC-BiMN), an adaptive hybrid neural framework, enhances feedback accuracy, learning adaptability, and instructional intelligence in product design education. Outlier detection identifies and removes anomalous interaction records to preserve meaningful learning patterns, while Min–Max normalization standardizes heterogeneous features for stable model training. Data augmentation is incorporated to enhance dataset diversity and reduce overfitting by generating representative variations of learner interaction patterns. Linear discriminant analysis refines key indicators to maximize class separability. Multi-kernel convolutional (MKC) layers extract diverse spatial behavior features, and a bi-directional memory module retains sequential interaction dependencies crucial for learning progression. DEE tunes network parameters to avoid premature convergence and improve optimization stability. Evaluated on learner interaction data and activity logs from AI-supported product design environments, DEE-MKC-BiMN achieves 96.86% of accuracy, demonstrating enhanced predictive performance, faster convergence, and improved adaptability compared with conventional hybrid models. DEE-MKC-BiMN strengthens intelligent tutoring effectiveness through multi-scale feature learning and fosters superior adaptive learning outcomes.

Graphical abstract