我国政务微博转发规模分类预测

The Retweeting Scale Classification Prediction of Government Microblogs in China

[目的/意义]政务微博是政府发布公共信息、与民众互动的重要平台,是创新社会治理的重要内容。因此,政务微博信息发布之后的转发规模对于预测网络舆情有重要意义。[方法/过程]从政务微博用户特征、内容特征、时间特征三个维度,提取特征体系,构建政务微博信息转发规模预测模型,用来刻画影响政务微博转发规模的政务用户权力层级、职能领域、传播形式等影响因素,通过比较多种机器学习算法测算政务微博转发规模分类预测的性能。[结果/结论]研究结果表明,随机森林算法在预测分类中表现最优;在对影响政务微博转发规模的特征重要性排序实验中发现信息发布时间、政务微博活跃度、社会影响力、政务微博行政级别等对预测精度的影响最为显著。

[Purpose/Significance]Government microblog is an important platform for the government to release public information and interact with the populace,and an important component of innovative social management.Therefore,the retweeting scale of the government microblogs is very significant for predicting the Internet public opinion.[Method/Process]From the three aspects of users'feature,content feature and release time feature of government microblogs,this study extracts a feature system for predicting the government microblog retweeting scale.In order to describe factors that influence the retweeting scale,such as the power hierarchy of microblog users,the government function areas,spreading forms,and so on.By comparing various machine learning algorithms,the study attempts to measure and calculate the properties of the classification prediction of government microblog retweeting scales.[Results/Conclusion]The results show that the Random Forest Algorithm performs the best in prediction classification.The feature importance ranking experiment finds that the most significant effect on prediction precision comes from information release time,microblog activeness,social influence,administrative rank of government microblogs,etc.