Stochastic processes on complex networks have a range of important implications in the real world. Yet, although the complete information of the network is often not present in cases of interest, there is no significant progress that would include this feature in prediction of stochastic processes. To address this issue, in this project, we will investigate how to predict stochastic dynamics on the networks with incomplete information. For this task we will further develop theory of approximate methods to predict some stochastic processes extending the existing methods to include computation of errors and to include cases of processes with non binary number of states. We will also develop a deep neural network to learn stochastic dynamics from the data with incomplete information of the network structure. These Data will be provided through extensive simulations of stochastic processes on computer cluster. Stochastic processes to be used will be carefully chosen and a special scheme to sample parameters of interest will also be developed. Furthermore, we will compare those two research directions and identify which approach works better in which cases. In the end of the project we should acquire enough knowledge to be able to construct prediction tools that could be used on real data in real world situations.