篇名 | Research on the Public Opinion Early Trend Prediction Based on the Trend Similarity of Emergency |
---|---|
卷期 | 31:2 |
作者 | Fu-Lian Yin 、 Xiao-Wei Liu 、 Bei-Bei Zhang |
頁次 | 101-113 |
關鍵字 | BP neural network 、 classified prediction 、 combination prediction 、 genetic algorithm 、 time series matching 、 EI 、 MEDLINE 、 Scopus |
出刊日期 | 202004 |
DOI | 10.3966/199115992020043102010 |
Under the environment of micro-blog communication in China, the current public opinion trend forecasting models can only forecast the trend of a complete single event, but the result of the early trend forecasting is very poor. To solve this problem, an early trend predicting algorithm based on the develop trend similarity of hot events is put forward in this paper. We apply the clustering results of the time series of hot events to build the prediction model for different categories, in which the GABP (genetic algorithm optimized BP neural network) is used to build the prediction model. We also propose the matching rules between a new event and the existing category. According to these, we realize the developing trend prediction for the new events. The experiment shows that the early trend forecasting algorithm proposed in this paper has better accuracy and timeliness when predicting the early development trend of public opinion. The APE (Absolute Percentage Error) of about 75% samples is below 80%, MSE (Mean Square Error) is below 0.01. And comparing with the traditional predicting algorithm, MSE is decreased by 90%, and APE decreased by 24%.