The approaches discussed in this paper undoubtedly make it possible to solve promptly the whole set of problems that objectively exist in the implementation of ICP, taking into account the factors of internal and external uncertainty. In fact, by implementing a machine learning model with reinforcement learning based on Markov decision process, we get a framework of decision support system in which the model itself can quickly analyze the progress of the project, build very flexible and accurate strategies with minimal errors, and such a model can include any internal contours: regression, classification, clustering, deep learning neural network models. As a result, the parameter of organizational and technological reliability of ICP significantly increase, which was required to obtain.