A Composited Framework for High-precision Network Public Opinion Risk Event Prediction

Predicting network public opinion risk events is important for the government's early reactions and public governance, but these events are usually uncertain and complicated. In order to increase the accuracy and comprehensiveness, and to reduce the prediction time such events, this work proposes a high-precision prediction framework of network public opinion risk. The framework first applies a four-stage particle model to represent the evolution process of network public opinion risk events. Then a correlation analysis method is adopted to screen the risk event prediction index data, and the rationality of the selected prediction indexes is verified. Moreover, the framework uses the geometric average weakening buffer operator to preprocess the initial data, so as to reduce the data distortion caused by the impact disturbance. Finally, the prediction results are displayed by the visual detection model. Experimental results on real data traces show that our framework has improved the prediction accuracy, comprehensiveness and prediction efficiency compared with tradition.