随着大规模在线开放课程（MOOC）的兴起，越来越多的学习者参与到在线课程当中。其中，在线学习环境下如何对学习者进行学习评估与学习成绩预测引起了研究人员的关注。在线学习环境下学生的学习行为与学习成绩密切相关，而学习分析技术为在线学习环境下的学习行为分析提供了技术条件。通过在线学习环境下学习者的学习行为预测学习成绩，能够帮助教师识别学习困难的学生并及时提供帮助。目前对于学习成绩预测研究多侧重于外显行为数据（如登录次数、观看视频的时长等）的分析，而对于反映学习者深层学习过程和认知活动的内隐行为数据的研究相对较少。在线学习平台的论坛讨论区，基于学习者产生的文本数据可以探究其隐含的情感、认知和知识建构等内隐行为信息。本研究以论坛讨论区学生产生的论坛讨论文本数据为研究对象，使用内容分析法、滞后序列分析法和社会网络分析法，从情感、认知和知识建构三个维度，深入挖掘高低成绩学习者之间的差异，以探究其中的关键信息。

本研究首先梳理国内外有关学习成绩预测模型、基于学习行为对学习成绩的预测和基于学习分析技术的学习行为的相关研究，明确内隐行为的定义。依据相关理论，基于相关学习分析模型框架，从情感、认知和知识建构三个维度来构建学习成绩预测模型。随后，基于MOOC讨论区文本数据进行实验研究，通过数据爬取技术获取讨论区文本数据，基于相应编码框架，通过学习分析技术探究高低成绩学习者的内隐行为模式差异。研究结果表明，对高低成绩学习者在线讨论文本的情感、认知和知识建构三个类别进行量化统计，部分内隐行为类别存在显著差异。高低成绩学习者在情感、认知和知识建构行为序列模式上存在不同的行为序列模式。同时，高低成绩学习者在在线讨论中，存在不同的社会网络结构。

为此，相关教师和学习平台可以有效的精准识别学习困难者并提供学习帮助，另一方面，本研究也为在线学习环境下的学习成绩预测提供了新的视角。

With the rise of Massive Open Online Courses (MOOC), an increasing number of learners are participating in online courses. In this context, how to assess learners and predict their academic performance in an online learning environment has garnered researchers' attention. Learning analytics technology provides the technical means to analyze learning behaviors in online learning environments. Predicting learners' academic performance based on their behaviors can help teachers identify students who are struggling and provide timely assistance. Currently, research on predicting academic performance often focuses on the analysis of explicit behavioral data, such as login frequency and video watching duration, with relatively less emphasis on implicit behavioral data that reflect deeper learning processes and cognitive activities. The forum discussion areas of online learning platforms, based on the textual data generated by learners, offer a means to explore implicit behavioral information such as emotions, cognition, and knowledge construction. This study uses the forum discussion text data produced by students as the research object and employs content analysis, lag sequential analysis, and social network analysis to delve into the differences between high and low performers from the dimensions of emotion, cognition, and knowledge construction, aiming to uncover the key insights.

This study first reviews the domestic and international research on models predicting academic performance, predictions based on learning behaviors, and studies related to learning behaviors using learning analytics techniques, thereby clarifying the definition of implicit behaviors. Based on relevant theories and the framework of existing learning analytics models, this research constructs a model for predicting academic performance, incorporating three dimensions: affective, cognitive, and knowledge construction. Subsequently, an experimental study was conducted using text data from MOOC discussion forums. Text data from the forums were collected using web scraping techniques. Using a corresponding coding framework, learning analytics were applied to explore the differences in implicit behavior patterns between high and low achievers. The results show significant differences in some categories of implicit behaviors between high and low achievers, particularly in the sequences of affective, cognitive, and knowledge construction behaviors. Furthermore, high and low achievers exhibit different social network structures in online discussions. Thus, this study not only allows educators and learning platforms to effectively and accurately identify learners who are struggling and provide necessary support but also offers a new perspective for predicting academic performance in online learning environments.