The problems of risk classification are of high importance in the current period of improvement of the system analysis methods for decision making in poorly structured and formalized problems from the complex systems of different nature. The paper considers the problem of developing a formalized and comprehensive description of risks in complex systems of different nature, particularly in engineering systems. Background is caused by the inability of modern approaches to consider risks complexly, compare different risks with each other, analyze the interdependence of risks and identify cascade effects etc. The proposed concept of risk classification is based on the formalized description of the risk. This approach allows the automating the process of risk classification on the basis of new data during the system running. Analytical classification of the risk is shown with the resource-based clustering of system resources that are at risk of loss. Clustering of the resources is performed on the closeness of values of the total cost of the system resources that are expressed in a single measure of resources. Analyzed risks classified and combined into one group, if they affect the resources that make up a single group (cluster). Analytical risk classification takes into account and complexly considers all risks of the system, allows predicting and controlling not only the individual risks but the whole risk clusters of the system. In addition, proposed approach enables to build the structure of risk classification for a particular problem, gives possibility to move from a verbal classification toward formalized one and ensures greater sustainability for risk management system. This concept of risk classification minimizes costs for risk management and the losses from negative effects for the entire system in general. The future research should be focused on the developing the approaches to risk classification on several parameters simultaneously.