In conditions of rapid external changes the requirement to quality of control of purposeful development of complex system (states, regions, corporations etc.) dramatically increases. Automation support of the key stages of decision making process is one of the ways to cope with the challenges. This paper focuses on the approach based on the Big Data Refining during cognitive modeling that proves the correctness of modeling and decision-making. The approach uses the requests to the Big Data for cognitive model components verification. The requests are created by intelligent agents with feedback from decision makers. Some practical results confirm the adequacy of the proposed approach.