Sentiment Analysis for Public Opinion Based on MapReduce and PSO-SVR

To conduct an emotional analysis of public opinion on campuses in a timely and effective manner, this study proposes a model for public opinion analysis and prediction based on a combination of hybrid PSO-SVR and MapReduce. This model is based on the SVR algorithm, which establishes a prediction model for emotional tendencies and degrees of public opinion. Utilizing the PSO algorithm to explore the optimal hyperparameter combination of SVR improves the reliability of the optimal hyperparameter combination and thereby enhances the effectiveness of sentiment analysis. In addition, using MapReduce to process massive amounts of public opinion data improves the accuracy and efficiency of the model. This experiment crawled TikTok comment data related to campus bullying events, cleaned and processed the data, used the PSO-SVR model for public opinion analysis, and compared the results with those of current popular machine learning models. The experimental results show that the PSO model can effectively find the optimal hyperparameter combination for SVR. In addition, the method based on the combination of MapReduce and PSO-SVR performs better than traditional models such as SVR, DTR and MLP.