基于回归模型与谱聚类的微博突发话题检测方法

发布时间：2018-01-17 17:21

本文关键词：基于回归模型与谱聚类的微博突发话题检测方法　出处：《计算机工程》2015年12期 　论文类型：期刊论文

【摘要】：微博社交网络短文本具有数据规模巨大、快速传播、模态多样、质量较低等特性,导致现有传统的话题检测与跟踪技术在对其进行数据处理时面临复杂度高、特征稀疏和噪声干扰等问题。为此,提出一种在回归预测和谱聚类基础上的突发话题检测方法。该方法针对关键词词频变动趋势,基于回归模型,量化微博关键词的突发程度,从词频趋势分析的角度准确地提取出突发词集合。设计一个基于谱聚类思想的突发词聚类方法提高聚类结果的准确性。在大规模微博数据集的实验结果证明,与baseline方法相比,该方法的准确率、召回率、F值都有较大提高,在微博信息分析领域有着较好的应用前景。
[Abstract]:Weibo social network short text has the characteristics of large data scale, rapid transmission, diverse modes, low quality and other characteristics, resulting in the existing traditional topic detection and tracking technology in its data processing faces a high complexity. Based on regression prediction and spectral clustering, a method of burst topic detection is proposed. The method is based on regression model for word frequency change trend of keywords. Quantifying the burst degree of Weibo's keywords. From the angle of word frequency trend analysis, the burst word set is extracted accurately. A burst word clustering method based on spectral clustering is designed to improve the accuracy of the clustering results. Compared with the baseline method, the accuracy of the method and the recall rate of F value are greatly improved, and it has a better application prospect in the field of Weibo information analysis.
【作者单位】：武汉大学计算机学院;武汉大学深圳研究院;
【基金】：国家自然科学基金资助项目“社会网络的主题演化分析与传播趋势预测研究”(61472291) 深圳市知识创新计划基础研究基金资助项目“基于压缩感知的社交网络主题提取与演化分析”
【分类号】：TP393.092;TP391.1
【正文快照】： 中文引用格式:彭敏,张泰玮,黄佳佳,等.基于回归模型与谱聚类的微博突发话题检测方法[J].计算机工程,2015,41(12):176-181.英文引用格式:Peng Min,Zhang Taiwei,Huang Jiajia,et al.Microblog Sudden Topic Detection Method Based onRegression Models and Spectral Clusterin

【相似文献】