随着社交网络规模的不断扩大，用户转发行为分析与建模面临着社交数据高度稀疏性、网络结构多元异构性和用户意图多样性等挑战。针对这些挑战，本课题在全面深入分析和总结用户转发行为的主要影响因素及对其进行量化表示基础上，通过对用户话题兴趣分布进行基于弱监督话题模型的推断，构建独立级联模型下的用户间动态影响力量化模型，使之能够有效地解决多元社交因素下的用户转发行为预测问题。主要的研究内容包括：(1)基于弱监督话题模型的用户兴趣抽取和表示方法；(2)基于独立级联模型的用户间动态影响力量化模型和表示方法；(3)基于矩阵分解的用户转发行为预测算法实现及其在消息传播中的应用。预期研究成果将为针对网络空间的用户转发行为分析提供理论基础，为大规模行为量化及预测提供算法框架，为网络空间中消息传播预测的应用提供高精度和高效率的实现技术，促进网络空间下用户行为大数据研究的技术创新和应用升级。

With the rapid development of social networks and diversity demand of users, the user retweeting behavior prediction faces with lots of challenges, including the severe data sparsity, network topology heterogeneity and user intents diversity and so on. In order to solve this challenge, this project proposal proposes a multipy social factors retweeting prediction model based on matrix factorization model by taking fully into account user topic interests and social influence between users. The main research tasks are as follows: (1) user interest extraction and representation based on weakly-supervised topic model; (2) quantitative and representation of dynamic influence among users based on independent cascade model; (3) user retweeting behavior prediction based on matrix factorization algorithm implementation and its application in message. It is expected that the research results will provide theoretical foundations for user behavior analysis in social networks and an algorithm framework for quantification and prediction of large-scale social behavior. The outcomes will also provide the implementation techniques of message propagation prediction for high accuracy and high efficiency, and promote technological innovation and application upgrade in cyberspace.