Abstract: Aiming at the problems that how to accurately and quantitatively reveal and model the  network structure characteristics and sentiment evolution patterns of cultural public opinion in online communication, we  proposed a dynamic evolution model for cultural public opinion based on the integration of complex network  and affective computing. The model took the  data from the Douyin platform as the core, realized quantitative expression of  the public opinion communication network by constructing a multi-dimensional weighted relationship matrix, designed a sentiment-weighted community clustering algorithm  to identify sentiment resonance, obtained the fused features of emotional polarity and communication structure within the network, and  introduced a temporal prediction model of evolution of  a gated recurrent unit network combined with an attention mechanism to achieve dynamic modeling of sentiment fluctuations and communication trends. Experimental results show that the proposed model effectively characterizes the communication structure and sentiment distribution characteristics of Jilin cultural public opinion in cyberspace, and its fitting accuracy and predictive results  are superior to those of traditional Hegselmann-Krause (HK) opinion dynamics models and susceptible, infected, recovered (SIR) communication models, verifying that the model has strong generali
zation capability and can  provide effective theoretical support for analyzing the digital dissemination patterns of culture.

Key words: cultural public opinion, complex network, affective computing, cultural public opinion evolution