一种基于时间戳的新闻推荐模型

A NEWS RECOMMENDER MODEL BASED ON TIMESTAMP

互联网的高速发展,使用户很难在"信息海洋"中找到感兴趣的新闻,如何为用户准确推荐满足其需求的个性化新闻已成为当前研究的热点和难点。为了改善新闻推荐系统的准确性,将时间戳信息引入到新闻推荐模型中。首先,利用分词工具对新闻标题和新闻内容进行分词,并引进时间加权函数来计算用户对单个分词的偏好;预测用户偏好时不仅根据用户自身的偏好进行预测,还使用改进协同过滤方法来预测用户偏好;最后,通过融合得到的偏好值对新闻进行推荐。实验结果表明,该模型不仅能提高新闻推荐系统的准确性,还缩短了模型构建的响应时间。

Rapid development of Internet makes it difficult for users to find the interested news from a information oceanr,. It has been the hot issue and challenge in current studies that how to accurately recommend the personalised news to users meeting their requirements. In the paper, we introduced the timestamp into news recommendation model in order to improve the accuracy of the news recommender system.First, we employed the word segmentation tool to segment the news titles and news contents into words, and introduced the time weighting function to compute the preference of users on individual word segmentation. When predicting users preference, we were not just based on the preference of users themselves, the improved collaborative filtering method was also applied in prediction. Finally, the news recommendation was achieved by integrating the derived preference values. Experimental results showed that the proposed model could not only improve the accuracy of news recommender system, it also shortened the responding time of model building as well.