基于XGBoost的网络舆情文章预警模型建立

The development of early warning model of network public opinion articles based on XGBoost

现有的舆情预警模型主要是以指标体系为依据来综合评判预警的等级。研究中指标体系的建立及量化是通过专家经验知识获得,具有较强的主观性,缺乏以数据驱动的指标验证研究。针对以上问题,本文为网络舆情新闻文章设计了一套全面可量化的预警指标,并使用特征筛选的方法,来验证所提指标对网络舆情文章预警的重要性,建立基于XGBoost的网络舆情文章预警模型,从而实现对舆情文章预警判定。该模型与现有用于舆情预警建模的SVM和决策树两个方法进行实验对比,结果证明该模型在各项性能指标上均高于其他两种模型,对网络舆情文章预警等级判定更准确。

The existing public opinion early warning model is mainly based on the index system to comprehensively evaluate the level of early warning.However,in the current research,the establishment and quantification of the indicator system are highly subjective.The indicators are obtained through expert experience and knowledge,and there is a lack of data-driven indicator verification research.In view of the above problems,this paper designs a set of comprehensive and quantifiable early warning indicators for network public opinion news articles,and uses the method of feature screening to verify the importance of the proposed indicators to the early warning of network public opinion articles.Besides,an early warning model of network public opinion articles is established based on XGBoost to realize the early warning of public opinion articles.We compare our model with SVM and decision tree methods for public opinion early warning modeling.The results show that the model performs better than the other two models in various performance indicators,and the early warning level of network public opinion articles is more accurate.