Detection of Negative Online Public Opinion among College Students Based on STOA Optimization Algorithm and Dissemination Trend Data

Abstract

Under the background of rapid advancement of information technology and the Internet, the impact of online public opinion on social stability and campus environment is increasingly significant, especially the negative online public opinion involving college students. To effectively detect and manage negative online public opinion, this study proposes a public opinion classification model that combines the sooty tern optimization algorithm with support vector machines. By adopting three schemes to improve the traditional sooty tern algorithm and combining with neural networks, a public opinion prediction model with temporal characteristics is designed. The test results showed that the accuracy, recall, F1 value, and resource consumption rate of the classification model were 92.86%, 93.75%, 93.82%, and 18.63%, respectively. The average calculation time was 9.1 seconds, which was the best among the compared models. When the target error values were 0.5 and 0.1, the average training times of the prediction model were 49 and 94, and the average error values were 0.1853 and 0.0725. The experiment shows that the detection system can quickly identify and locate negative public opinion in large-scale data, and accurately predict its dissemination trend, providing a new warning information and response strategy for relevant management departments.