As the main body of the spread of information, user group receives wide attention from scholars with the booming development of mobile social network in recent years. In order to provide references for relevant departments so that they can effectively supervise the dissemination of public opinions, this paper studies the spread of public opinion based on the behavior of user group which is integrated to optimize SIR model. Then this study respectively constructs a dynamic evolution model of mobile social network public opinion dissemination in homogeneous and heterogeneous networks by combining replicator dynamics equation and it utilizes Jacobi matrix to analyze stability. Finally, this study checks the numerical and model simulation. It is observed that the scale and speed of propagation of public opinion is the same as those in both homogeneous and heterogeneous network after taking the behavior of user group into consideration. When the propagation rate is higher than the immune rate, the rising speed of propagation node ratio is faster and the peak value of it is higher. After considering user group, the peak value of the propagation node ratio of public opinion is higher and it has longer holding time. And the lager user group has a faster propagation rate of public opinion and a wider sphere of influence.