The paper proposes an approach to the early detection of factors implying the need in production schedule update. Resource state prediction methods are based on the development of a binary model and a machine learning techniques called association rules search.