一种融合用户关系的自适应微博话题跟踪方法

摘要

针对微博口语化、文本短小等特点以及现有研究的不足,本文提出了一种融合用户关系的自适应微博话题跟踪方法.首先,在当前跟踪的时间窗内,推文被映射到特征空间,并作为候选推文集合.然后,针对推文的分布特点以及话题跟踪的目的,变换推文特征空间.在此基础上,利用改进的K-means聚类算法对候选推文集合进行二元聚类,从而划分出相关推文集合,即当前话题目标模型.本文通过Twitter平台获取数据进行实验,实验结果表明,该方法能够实时地跟踪话题热度的变化以及焦点的演变,并提高了微博中话题跟踪的稳定性.该方法为用户推荐、舆情分析等领域提供了有效的支撑.%Considering the colloquial,short text and other characteristics of microblog and deficiencies in research of it,this article proposes a self-adaptive topic tracking method of microblog by user relationship.First of all,during the tracking time window,the candidate tweet set is mapped into feature space.Secondly,aiming at the characteristic of tweet distribution and the purpose of topic tracking,the paper converts the tweets'' feature space.Based on this operation,a binary clustering on tweets set can be constructed by improved K-means clustering algorithm.The yielded relative collection is the target model of the current topic.The experiments with the data extracted from Twitter,show that this method can track down the trend of hot topics and the evolution of focuses in real time,and improve the stability of topic tracking in microblog.This method serves well for user recommendation and public opinion analysis.