Emotion Prediction and Analysis of Weibo Users Combined with Portraits

Since the outbreak of the novel coronavirus, Weibo has become one of the important platforms for Chinese netizens to receive information related to the epidemic. Behind the complexity of the news, netizens' emotions often affect the general social atmosphere, and the rapid spread of extreme negative emotions is not conducive to social harmony and stability. Therefore, it is particularly important to predict the emotional tendency of netizens and to pay attention to and guide users with negative tendencies. To improve the accuracy, reduce the amount of calculation and improve the running speed, this paper proposes a prediction process based on the BERT + LightGBM model. By taking advantage of the respective advantages of the two models and combining the emotional data of microblog content and user characteristics, this paper realizes the emotion analysis and prediction of Weibo users. The validity of the BERT + LightGBM model was verified by the case of the "granddaughter of a confirmed case in Chengdu". Compared with BERT, LSTM, and CNN, the BERT + LightGBM composite model has higher accuracy and better application prospects.