微博个性化信息流推荐研究

Personalized tweet recommendation

针对为微博用户推荐符合其兴趣和喜好的个性化微博信息的问题,结合协同过滤的思想,基于TF-IDF模型综合考虑了单个词语向量和多个词语向量相结合的特点后,用于计算微博信息流的相似性并评估用户的兴趣度。通过进一步分析用户的冷启动的问题和个性化特点,有效降低了无关微博信息的排名,优化用户微博信息排序。将基于新浪微博数据集与现有的余弦相似性和标签向量的微博推荐方法进行了对比实验,实验结果表明,该算法的有效性。

To solve the problem of recommending useful tweets that match users＇ interests and likes effectively,an approach based on term frequency-inverse document frequency （TF-IDF） was proposed.To capture personal interests,the approach improved the TF-IDF model by combining the vectors of single terms and pairs of terms and evaluating the similarity between the set of user＇s tweets and the stream of tweets to users based on the idea of collaborative filtering.Moreover,the model studied the cold-start problems and personal features of users to optimize the queue of the tweets received by users.The experiments on the SINA BLOBS data showed that the proposed method could reduce the ranks of irrelevant tweets effectively and achieve better performance than several baseline methods based on cosine and hash tags.