基于传播特征的微博流行度预测算法研究  

Research on Weibo Popularity Prediction Algorithm Based on Propagation Features

作　　者：胡俊睿 邹海涛 郑尚[1] 于化龙[1] 高尚[1] HU Junrui;ZOU Haitao;ZHENG Shang;YU Hualong;GAO Shang(School of Computer,Jiangsu University of Science and Technology,Zhenjiang 212003)

机构地区：[1]江苏科技大学计算机学院,镇江212003

出　　处：《计算机与数字工程》2023年第4期763-768,964,共7页Computer & Digital Engineering

摘　　要：微博的传播特征通常比较复杂,已有的微博流行度预测研究或者是基于分类算法,其模型对于流行度的定义较为主观,或者仅考虑微博转发速度对其流行度的影响,模型的构建仍存在很大的改进空间。论文引入传播加速度的概念,将微博的转发加速度、回复加速度以及用户粉丝数等作为影响微博流行度变化的指标,把目标时刻的转发数、回复数作为衡量流行度的指标,通过观测微博传播的前K小时流行度变化情况,构造线性回归函数模型,预测目标时刻的微博流行度。通过测试微博的转发加速度、回复加速度以及用户粉丝数对模型的不同影响,实验结果表明论文提出的微博流行度预测模型性能良好,与传播特征数量关系密切,其预测结果优于仅考虑微博转发速度的预测模型。Existing researches on Weibo popularity prediction are either based on classification algorithm,in which the definition of popularity is relatively subjective,or it only considers the influence of the forwarding speed of Weibo on its popularity.In this paper,the concept of propagation acceleration is introduced,and the forwarding acceleration,the reply acceleration and the number of users'followers are taken as indicators to influence the variation of Weibo popularity and the number of forwarding and replying at the target time is taken as indicators to measure Weibo popularity,and a linear regression function model is constructed to predict the microblog popularity at the target moment by observing the popularity changes during the first K hours of Weibo propagation.By testing the different influences of the forwarding acceleration,the reply acceleration and the number of users'followers on the model,the experimental results show that the prediction model of Weibo popularity proposed in this paper has good performance and is closely related to the number of propagation features,and its prediction results are superior to the prediction model that only considers the forwarding speed of Weibo.

关 键 词：流行度预测 线性回归 加速度 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]