对于信息推送的内容过滤策略进行改进,在现有基于正例和无标注样例(PU)的学习理论的分类基础上,通过对反例文档的发现进行研究,将这种学习理论在信息推送中的内容过滤进行实验,通过实验证明整套策略在内容过滤上精度和速度都有明显的提升。
This paper aimed at promoting the method of content-filtering for information push. It focused on researching the method of finding negative examples based on the traditional positive and unlabeled examples learning theory. At last it took an experiment on text categorization. The experiment shows that this method is ideal for precision and speed of content filtering.