[Objective] Based on the analysis of online public opinion events, determining their attribute characteristics and classification. When a new online public opinion event occurs, we can predict whether the event will reverse in advance, which can not only help the government adjust the direction of public opinion in time but also prevent the credibility of the government media from being negatively affected.[Methods] The representative online public opinion events were collected in the past five years. The improved SMOTE algorithm made a balance distribution treatment on the event data set, building the prediction model of online public opinion reversal based on the neural network ensemble learning, and using the indicators such as accuracy rate, recall rate to evaluate the prediction effect of the model. The latest online public opinion events in 2020 are selected to test the proposed model, so as to further reveal the internal mechanism of the constructed reversal prediction model. [Results] Through empirical research, the accuracy of the neural network ensemble learning classification model is 99% and the values of F and AUC are both 0.99, which verifies the feasibility and strong generalization performance of the model.
[Limitations] This paper only selects some characteristics of public opinion reversal events for research. Therefore, it cannot comprehensively represent all public opinion reversal events that will occur in the future. [Conclusions] The constructed public opinion reversal prediction model can accurately predict whether the public opinion event will reverse in advance.
王楠, 李海荣, 谭舒孺.
[J]. 数据分析与知识发现, 10.11925/infotech.2096-3467.2020.0838.
Wang Nan, Li Hairong, Tan Shuru.
Research on prediction of public opinion reversal based on improved SMOTE algorithm and ensemble learning
. Data Analysis and Knowledge Discovery, 0, (): 1-.