针对当前建构“以学生为中心”的参与式课堂、提升高等教育内涵建设的重要需求与缺乏大规模、实时反馈的课堂参与度评价手段之间的矛盾，本课题提出场景数据驱动的参与度实时智慧评价分析框架，通过构建不同学科与类型课程中学生参与度的优化模型，为教师的课堂组织提供更为及时的针对性的反馈。与现有方法相比，本课题研究的优势体现在:1) 针对课堂参与度评价开展的各个阶段，在经典教育学模型的指导下提出数据驱动的采集、量化和评价方案，充分考虑人工智能技术与评价需求的结合方式，降低人力和时间开销;2) 针对课堂参与场景的普通特征和专业特征，提出课堂多媒态数据的自动提取和融合方法，覆盖教育学经典评价模型的大部分观测指标，并保持提取过程和结果的可解释性;3) 针对经典教育学量表中的定量评价和定性评价，提出基于深度学习的模型校准和训练方法，通过迁移学习、“视觉-文本”特征转换等多种技术，保持分析预测模型与专家量表的一致性。

Aiming at the inefficiency and subjective limitations in the current process of study participation evaluation, this research proposes a study scenario data-driven participation evaluation and analysis framework, which maps various technical evaluation methods to unified evaluation indexes through hierarchical multi-index system. As while a complete observe indicator representation method is designed to implement data analysis and calibration according to the characteristics of each studying scene case. Compared with the existing similar methods, the advantages of this research are listed as follows: 1) Aiming at all stages and aspects of classroom participation evaluation, a data-driven collection, quantification and evaluation scheme is proposed under the guidance of classical pedagogical theory, which takes full account of the combination of artificial intelligence technologies and evaluation needs, and uses algorithms instead of general manually data collection to reduce manpower and time; (2) According to the general and professional characteristics of classroom participation scenarios, an automatic feature extraction and fusion method of classroom multi-media data is proposed, which can cover most of the observation indicators of classical pedagogical evaluation index, and maintain the expandability of the extraction process and results; (3) For quantitative and qualitative evaluation of classical pedagogical evaluation index, a deep learning based modeling and collaborating method is proposed. The trained prediction model can maintain the output consistency with the pedagogical experts’ results through transfer learning, visual-text feature transformation and other AI techniques.