Construction of a Public Opinion Analysis and Revenue Prediction Model for Enterprises Based on KOL Factors

With the Internet growing at a quick pace, enterprises now operate in a more complex network environment where public opinion can have a significant negative impact on their sales, profit, market trust, and brand image. This study uses Naive Bayes to handle online public opinion, and lexical sentiment tendency and centering are employed to produce the sentiment score. The sentiment score is used to build the autoregressive distribution lag model (ARDL) based on the autoregressive model for historical operating income projection. The KOL influencing elements are then included in the model to further optimize it for accurate prediction. The study's findings demonstrate a considerable improvement in the model's prediction accuracy as well as a strong optimization effect, as the final model's MAPE values in the two firms drop from 16.21 and 18.23 to 9.42 and 10.66, respectively. KOLs have a substantial impact on the dynamics of public opinion and market sentiment. By including KOL factors and sentiment scores in the model, the model can better reflect market reality and improve the fit and accuracy of revenue prediction.