【摘要】 随着学习者个性化需求的增加和个体差异性的扩大，学习需要解决适应千人千面的个性化学习推荐问题。传统推荐算法主要面向欧式空间数据建模，忽略了现实中的图结构数据。从图神经网络角度出发，应用图卷积方式分别提取用户社交图结构下的用户特征数据，以及知识图谱结构下的知识特征数据，从而形成更加有效的学习推荐内容，实验数据也证明了算法的有效性。更多还原

【Abstract】 With the increase of learners’ personalized needs and the expansion of individual differences, learning needs to address the problem of adapting personalized learning recommendations. Traditional recommendation algorithms are mainly oriented to Euclidean spatial data modeling, ignoring the realistic graph structure data. Applying graph neural networks, graph convolution is used to extract user feature data under user social graph structure and knowledge feature data under knowledge graph structure, respectively, so as to form more effective learning recommendation contents, and experimental data also prove the effectiveness of the algorithm.更多还原