中文摘要：

推荐系统作为解决信息过载问题的关键技术，已经引起了国内外研究学者的广泛关注．迄今为止，业界最广受好评的是协同过滤推荐技术．但由于其本身存在着数据稀疏、冷启动等固有问题，而难以应对膨胀的社会网络这一应用场景．本文针对推荐系统所面临的挑战，构建合适的动态信任传递模型，所设计的基于信任的推荐算法是对稀疏性、冷启动等问题的有效解决方案，且对恶意攻击具备一定的抵抗能力．最后在真实社交网络数据中对所设计的算法进行实现，并与传统推荐算法做实验对比，实验结果表明算法相比协同过滤算法在准确性和覆盖率上表现更好，且算法具备的分布式特性在复杂社会网络与大数据环境下实现了推荐实时性的要求．

英文摘要：

As an important technologyto solve the information overloaded problem, recommender system has gained more and more at- tention from both domestic and internationalscholars. So far, the best recommender system of industry uses the collaborative filtering recommendation technology. Collaborative filtering algorithm itself, however, has inherent problems, such as data sparsity, cold start. Hence, it is difficult for a simple recommendation mechanism based on traditional collaborative filtering algorithm to deal with the ex- pansion of the social network as the scenario. I builds appropriate dynamic trust modeland design an efficient distributed recommenda- tion algorithm, which is a good solution to the data sparsity and cold-start issues and anti-attacking. In the last, I implement the algo- rithm and experiment in the real social network data sets, and compared with the traditional recommendation algorithm. It turned out to better in accuracy and coverage, and the distributed manner fits the real-time requirement with the background of complex social net- work and big data.