基于偏差规则马尔可夫模型的网络舆情情感预测研究

Emotion Prediction of Network Public Opinions Based on the Deviation Rules Markov Model

针对微博短文本评论,基于情感分析技术,从情感类转移等角度实现对网络舆情演变趋势的预测。本文以突发事件“新冠肺炎疫情”初期的相关微博评论文本作为研究对象,基于扩展关联规则Apriori算法和马尔可夫链提出一种新的方法,即偏差规则马尔可夫模型(the deviation rules Markov model,DRMM)。该模型分析了网民情感类间的相关性和转移性,通过计算不同的情感类转移概率、构建时变的情感状态转移矩阵对疫情初期网民情感状态的变化趋势进行预测。实验采取平均绝对误差(mean absolute error,MAE)和均方根误差(root mean squared error,RMSE)来衡量模型预测值与真实值之间的误差。研究结果表明,该模型具有较好的有效性和准确性,预测值和真实值的拟合效果在预期范围之内。

Using sentiment analysis technology,the evolution of online public opinion can be predicted from the perspective of emotion transfer in short microblog comments.In this study,the short texts of related microblog comments in the early stage of the COVID-19 pandemic were taken as the research object.Based on the extended association rule Apriori Algorithm and Markov Chain,a new method called the deviation rule Markov model is proposed.This model analyzes the correlation and transfer between Internet users’emotion classes and predicts the changing trends of Internet users’emotional states in the early stage of the pandemic by calculating the transfer probability of different emotion classes and constructing a time-varying emotion state transfer matrix.The experimental results demonstrated that the emotional state of netizens after the pandemic outbreak was not negative but gradually changed to“positive”emotions over time.Through a comparative analysis of examples,the validity and accuracy of the affective prediction model for online public opinion proposed in this study were verified.