Prediction of Negative Public Opinion Dissemination Trend on Social Media: An Improved Machine Learning Algorithm

Abstract

The widespread dissemination of negative public opinion on social media can have adverse effects. This study utilized word-to-vector (word2vec) to obtain word vectors and classified positive and negative public opinions using the bilateral long short-term memory (BiLSTM) algorithm. To predict the dissemination trend of negative public opinion on social media, the back-propagation neural network (BPNN) algorithm was selected and optimized by the improved sparrow search algorithm (ISSA). Public opinion data was collected from the “Wenzhou doctor injury” case for experiments. The results showed that the BiLSTM algorithm achieved a P-value of 0.9233, an R-value of 0.9164, and an F1-value of 0.9198 in sentiment classification, outperforming the convolutional neural network and long short-term memory algorithms. For Benchmark test functions, the ISSA demonstrated superior performance in optimization compared to the particle swarm optimization and sparrow search algorithm. In predicting the negative public opinion dissemination trend, the ISSA-BPNN algorithm yielded a mean square error of 845.12, a root-mean-square error of 29.07, and a mean absolute error of 21.56, surpassing support vector machine and other algorithms. These results validate the effectiveness of the method proposed in this study for predicting the dissemination trend of negative public opinion and its potential practical applications.