论文部分内容阅读
为解决社交媒体中缄默用户的性别预测问题,提出利用用户文件夹中的兴趣标签进行区分的方法.针对标签存在稀疏和歧义性的特点,设计了一种基于概念类推断用户性别的框架.首先依据社交心理特征将标签划分为若干概念类;其次通过关联挖掘方法扩充概念类;最后通过概念类压缩用户特征空间.在新浪微博真实数据集上进行验证,实验结果表明:所提方法对于缄默用户性别有显著的区分效果,在不使用任何微博信息的条件下,区分准确率达到71%.
In order to solve the problem of silence prediction for social media users, this paper proposes a method to distinguish the interest tags in the user’s folders.Aiming at the sparsity and ambiguity of the tags, a framework for inferring user’s gender based on concept classes is proposed. According to the social psychological characteristics, the tags are divided into several conceptual classes. Secondly, the concept classes are extended through the association mining method. Finally, the concept classes are used to compress the user characteristic space. The verification is performed on the real data set of Sina Weibo. The experimental results show that: User gender has a significant distinction between the results, do not use any Weibo information conditions, the distinction between the accuracy of 71%.