Research on the Application of EMD Decomposition-based AR Model in Popularity Prediction of Public Sentiment on Events

Under the background of continuous development of the network, more and more people tent to voice their opinions on the network, which makes the network public sentiment gradually develop into the main battlefield of public sentiment, and also makes the influence of network public sentiment on events expand. The development trend of most events in nowadays can be detected and predicted from the network public sentiment and make relevant analysis and prediction. However, for the complex network public sentiment information, the traditional single time sequences model has the characteristics of low accuracy and cannot achieve the demand of accurate prediction. Moreover, the original public sentiment time sequences data does not demonstrate obvious regularity and predictability, which undoubtedly increases the difficulty of public sentiment prediction. Therefore, in order to improve the accuracy of the model, an autoregressive model based on empirical mode decomposition is proposed in this paper. Firstly, the original public sentiment event sequence is decomposed into several stable connotative modal components and a residual component by empirical mode decomposition and data difference processing method. After being smoothed by the unit root test, the autoregressive model is applied to predict each order component separately, and finally the reconstructed results are compared with the results of the autoregressive model alone. The results show that the prediction accuracy of this method is significantly improved as compared to that of the AR model, indicating that this method is more effective in making predictions for time sequences public sentiment data.