基于信息传播理论的微博协同过滤推荐模型

高速发展的微博带来信息富余,也带来了信息过载,不断新增的非结构化微博文本内容和复杂的社会网络关系导致个性化推荐难以实施.针对微博网站特征,提出一种基于信息传播模拟的协同过滤推荐模型并给出推荐框架图,解决推荐的数据稀疏性和冷启动问题.首先,通过自然语言处理技术处理非结构化文本内容,获取关键词为推荐对象,构建用户-关键词偏好模型;然后,采用一阶马尔可夫随机游走模拟用户偏好在社会网络中的传播过程,得到用户-关键词偏好矩阵.实验使用来自新浪微博的数据集,采用平均绝对误差、准确率和召回率三个指标评价推荐模型,并与基准模型进行对比.实验结果表明,因整合了社会网络结构信息,基于信息传播的协同过滤推荐模型的效率比基准模型有明显提高.

Collaborative filtering recommendation model in micro-blogging website based on information diffusion theory

With the fast development of micro-blogs, there is an unprecedented abundance of information shared online. Meanwhile, information overload appears. Due to the continually increasing nonstructural contents and the complexity that characterizes the structure of social networks, the deployment of online customized recommendation systems is very challenging. In this article a collaborative filtering recommendation model integrating information diffusion has been designed to solve the problem of data sparsity and cold-start problem in the context of micro-blog websites. Firstly, the proposed framework constructs a user-keyword interest model by using natural language processing technology to handle nonstructural contents.extracted thereafter act as recommended items. Secondly, the research uses a first-order Markov random work model to simulate the processing of users' preference' diffusion in social networks. The user-keyword preference matrix is then conducted. Lastly, the experiments are conducted with real world dataset from Weibo. The effectiveness of the proposed recommendation model is evaluated by using three assessment criteria (mean absolute error, precision and recall). The results show that the collaborative filtering recommendation model integrating information diffusion performs better than benchmark models which do not consider social networks.