Research On Sentiment Analysis And Evolution Trend Prediction Technology Of Weibo Public Opinion Under Emergencies

In recent years,various major unexpected events have occurred frequently,accompanied by the continuous increase of internet users and innovative development of intelligent terminals.The data collection and analysis based on social networking platforms such as Weibo for unexpected events have been widely concerned.It is crucial for the relevant regulatory agencies to research the formation and development of online public opinion in the mobile environment during unexpected events,accurately understand the emotional characteristics of online public opinion in the mobile environment,and grasp the trend of public opinion evolution.Therefore,this paper proposes a four-step method for sentiment analysis and trend prediction of Weibo public opinion during sudden events,focusing on research from two aspects: sentiment analysis and trend prediction of public opinion evolution.As public opinion data continues to increase and various major unexpected events occur,Weibo text sentiment analysis technology has become increasingly important in monitoring online public opinion.Due to the sparsity and high dimensionality of public opinion text data,as well as the complexity of natural language semantics,sentiment analysis tasks during sudden events face significant challenges.In this paper,based on the BERT word vector model and deep learning model,an external sentiment lexicon is introduced to enhance the emotional color of the text.Specifically,this study first uses BERT to dynamically represent the text with word vectors,and uses a pre-processed sentiment lexicon to enhance the emotional features of the text.Then,a Bi LSTM network is used to extract the contextual features of the text,and an attention mechanism is used to weight the processed vectors,followed by a CNN network to extract the main emotional features of the text.Finally,the processed emotional feature representation is classified.Comparative experiments on a public dataset of a certain sudden event show that the proposed model has a significant improvement compared to other similar models.In terms of predicting the trend of public opinion evolution,considering the difficulty in obtaining data during sudden events,this paper uses the grey prediction model with good predictive ability under sparse information to replace traditional prediction models for predicting public opinion evolution trends.However,due to the high volatility and timeliness of public opinion data generated during sudden events,conventional grey prediction models are difficult to predict accurately.Therefore,this paper proposes a novel SISGM(1,1)model that optimizes both background value and initial condition.The model uses Simpson’s formula to adjust the background value,which can better smooth data sequences and handle data with high fluctuations.In addition,setting the ISRU activation function as the initial condition of the grey prediction model can ensure the priority of new information principle and make the time response function approach most functions,thereby increasing the adaptability of the model.Finally,the particle swarm algorithm is used to calculate the introduced parameters to further improve the prediction accuracy.Experimental results show that the proposed model has significant advantages over other competitive models,demonstrating the effectiveness of the model.Finally,based on the above-mentioned sentiment analysis and trend prediction models,this paper will conduct a systematic exploration and research on the proposed four-step method for sentiment analysis and trend prediction of Weibo public opinion during sudden events.Through concrete examples,the feasibility of this method will be verified.