The problem of detecting errors and improving a complex large-scale project is considered as the task of data mining. A model of selective research of such project is designed. The task is to select the information files of the project, the expertise of which ensures the greatest achievement of the goal (for example, the maximum reduction in the cost of the project). Its decision is based on the principle of equal job complexity: modules from files studied by specialists of the same qualification should be equal in complexity. A heuristic algorithm for such a solution is described. The problem of estimating the necessary number of experts was considered and solved on the basis of an assessment of the overall job complexity, taking into account the costs of their coordination. Based on the results obtained and the principles of data mining, an approach to the examination of complex large-scale projects, illustrated by examples in the field of transport, was developed.