基于动态特征的学者推荐研究

Research on Scholar Recommendation Based on Dynamic Features

[目的/意义]在开放的知识交流环境中向学者推荐具有相似研究兴趣的学者有助于学者高效获取到所需的学术资源,更好地开展学术研究和学术交流。[方法/过程]首先利用LDA主题模型提取学者的科研兴趣特征,同时引入时间因子加权兴趣特征,形成学者动态兴趣矩阵,基于此使用K-means对具有相似研究兴趣的学者进行聚类分析,并在类簇内综合学者的科研能力和社交属性两个维度构建学者推荐模型。[结果/结论]以“百度学术”数据集对模型进行验证,实验结果表明该模型能够很好地发现相关学者,满足可操作性和推荐结果有效性。在学者推荐过程中引入更贴近现实的动态兴趣特征对推荐结果具有一定效果。

[Purpose/significance]Recommending scholars with similar research interests to scholars in an open knowledge exchange environment will help scholars to efficiently obtain the required academic resources and better carry out academic research and academic exchanges.[Method/process]Firstly the LDA topic model is applied to extract the scholars’scientific research interest features,and the time factor weighted interest features are introduced to construct a scholar’s dynamic interest model,and then K-means is used to cluster scholars with similar research interests,and a scholar recommendation model is constructed by integrating the two dimensions of scholars’scientific research ability and social attributes in the cluster.[Result/conclusion]The model proposed in this paper is verified with the dataset of“Baidu Academic”.The experimental results show that the model can find relevant scholars well,meet the operability and the validity of the recommended results.The introduction of more realistic dynamic interest features in the process of scholars’recommendation has a certain effect on the recommendation results.