The purpose of this research is to estimate the probability of message dissemination when data is incomplete. To achieve this, diffusion models and their application in estimating message transmission probabilities have been investigated. We proposed a method based on the dynamic message passing algorithm and graph neural networks. This method allows for the estimation of message transmission probabilities. Importantly, it is independent of any prior knowledge about the incomplete parts of the observations. This method infers the rate of message transmission between vertices with a lower error compared to previous methods.