将学习者学习风格的偏向性整合进智能学习环境中的研究是一个处于计算机、心理学、教育学领域交叉中心的研究热点。但是现有智能学习推荐系统均存在学习风格预测模型准确率低以及基于学习风格用户模型的资源推荐模型性能差等问题。本课题对此展开研究，主要内容包括：(1)动态学习风格预测模型研究，包括认知受限多标签分类、多维度特征空间提取、动态聚类等；(2)跨学习风格分类族的学习资源推荐研究，包括学习资源的自组织、跨学习风格分类族的学习资源推荐模型以及学习行为引导等。本课题拟采用基于相互相似关系模式识别的方式来强制将多标签分类问题转化为单标签分类问题；采用自动编码器进行多维度特征空间提取；采用加入样本可靠性检测的动态聚类算法对学习者进行聚类；采用改进的多向SLN来建立可演化的学习资源自组织结构；采用学习资源推荐推理树、推理规则动态排序和路径推荐规则来实现跨学习风格分类族的资源和学习路径推荐。

How to integrate the learning style (LS) preference of the learners into a personalized e-learning environment is an important research issue at the intersection of computer, psychology and education. The existing intelligent learning systems have two main problems when integrating LS profiles into their systems, which are: the low accuracy of the LS prediction model and the low efficiency of the recommendation model based on users’ LS profiles. This project intend to resolve the aforementioned problems by study the following critical subjects: (1) to dynamically predict the users’ LS preferences by using mutual similarity pattern recognition technique on solving the cognitive constrained multi-label classification problem, using AutoEncoder on filtering learning behavior features in different LS dimensions and using reliable checking component on clustering; (2) to effectively recommend learning resources and learning paths based on LS profiles by using the self-organized learning resources, conflicting resolving mechanism and learning path recommendation rules across different LS families.