The construction of county-level integrated media is facing a communication dilemma, and it must alter the way and method of communication by using big data. In this paper, under the background of big data, the basic metric features of the users of integrated media communication are calculated by name similarity, one-hot coding features, following rate, etc., which are combined with the pixel statistical characteristics and avatar personality features to build a user portrait of integrated media communication. Taking the user’s profile as the anchor point of communication, the CD-VAE model is used to predict and analyze the early communication of integrated media content. Through case studies, the current characteristics of county-level integrated media and the trend of primary and secondary dissemination are examined. The first and second newspapers were distributed 278 times and 251 times, respectively, with a conversion rate of 12.61%. The primary and secondary dissemination of Jitterbug reached 303 and 284 times, respectively, as media in the era of mobile Internet. The use of big data technology can effectively analyze the development pain points of county-level integrated media.