Scholarlink

<p>Teaching mode refers to the method used to deliver educational content, including lectures, demonstrations, and hands-on activities, designed to meet students’ diverse needs and engagement. Teaching mode innovation is transforming education by utilizing advanced technologies to create dynamic learning experiences that cater to the diverse needs of students. Deep, Reinforcement Learning (DRL) has been explored for adaptive learning. Existing approaches struggle to effectively utilize a large amount of student performance data due to delayed and sparse feedback, reliance on static data, and a lack of real-time adaptivity. These issues prevent timely adjustments to teaching strategies, limiting their ability to optimize learning experiences. Additionally, existing research models face the challenge of delayed and sparse feedback, which prevents timely adjustments to teaching strategies based on students’ real-time data, thereby limiting the effectiveness of learning and engagement among students. The research explores the application of Reward Shaped Double Deep Q-Learning Network (RS-DDQN) integrated with big data mining techniques to adapt to innovations in teaching modes. The RS-DDQN addresses the issue of delayed and sparse feedback by utilizing reward shaping. This technique enhances the feedback signals, making them denser and more informative. It helps the model learn faster by providing more immediate and valuable feedback, enabling real-time adjustments to teaching strategies, even when feedback is infrequent or delayed. The proposed method involves applying RS-DDQN to analyze large-scale educational data, which comprises student performance metrics, learning patterns, and interaction data, to optimize the teaching process dynamically. The model processes multimodal educational data, including student performance indicators and engagement patterns, across different teaching modes such as lectures, demonstrations, and hands-on activities. The experimental results have been analyzed over 15 weeks, involving 13 student groups, for various scenarios. The results demonstrate significant improvements in teaching effectiveness, with a 27% increase in student engagement, a 90% reduction in training complexity, and an 11% enhancement in adaptive teaching assessment compared to existing models.</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

Adaptive teaching mode optimization using reward-shaped deep reinforcement learning and big data mining

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年1月13日

查看全文
  • Hong Cheng,
  • Ber-Lin Yu

摘要

Teaching mode refers to the method used to deliver educational content, including lectures, demonstrations, and hands-on activities, designed to meet students’ diverse needs and engagement. Teaching mode innovation is transforming education by utilizing advanced technologies to create dynamic learning experiences that cater to the diverse needs of students. Deep, Reinforcement Learning (DRL) has been explored for adaptive learning. Existing approaches struggle to effectively utilize a large amount of student performance data due to delayed and sparse feedback, reliance on static data, and a lack of real-time adaptivity. These issues prevent timely adjustments to teaching strategies, limiting their ability to optimize learning experiences. Additionally, existing research models face the challenge of delayed and sparse feedback, which prevents timely adjustments to teaching strategies based on students’ real-time data, thereby limiting the effectiveness of learning and engagement among students. The research explores the application of Reward Shaped Double Deep Q-Learning Network (RS-DDQN) integrated with big data mining techniques to adapt to innovations in teaching modes. The RS-DDQN addresses the issue of delayed and sparse feedback by utilizing reward shaping. This technique enhances the feedback signals, making them denser and more informative. It helps the model learn faster by providing more immediate and valuable feedback, enabling real-time adjustments to teaching strategies, even when feedback is infrequent or delayed. The proposed method involves applying RS-DDQN to analyze large-scale educational data, which comprises student performance metrics, learning patterns, and interaction data, to optimize the teaching process dynamically. The model processes multimodal educational data, including student performance indicators and engagement patterns, across different teaching modes such as lectures, demonstrations, and hands-on activities. The experimental results have been analyzed over 15 weeks, involving 13 student groups, for various scenarios. The results demonstrate significant improvements in teaching effectiveness, with a 27% increase in student engagement, a 90% reduction in training complexity, and an 11% enhancement in adaptive teaching assessment compared to existing models.