Cost Management with a Standard and Two Self-Learning Rates

When controlling costs in industrial systems, it is necessary to take into account the asymmetric awareness of their elements. It provokes undesirable activity of these elements. We study a two-level active system of cost control. At its top level is the Overseer, and at the bottom is the cost management entity, which is directly responsible for the costs. To minimize costs, the Overseer uses standardization, machine learning, normalization and classification procedures to assign management to one of the four classes. The better the class, the higher the incentive for cost management entity. To solve the problem of synthesizing such a hybrid mechanism, a decomposition is applied. First, an archetypal mechanism with machine learning, normalization, and binary classification is developed, taking into account the activity of the cost management entity. It is proved that the archetypal mechanism minimizes costs and improves the efficiency of machine learning and normalization. Second, a hybrid mechanism is designed, including a standardization procedure and two archetypal mechanisms (with machine learning, normalization, and binary classification procedures) for assigning four classes to the cost management entity. However, at some values of initial rates, such a hybrid mechanism can lead to a lack of progress in reducing costs and stagnation. When this problem occurs, it is sufficient to restart the hybrid mechanism, choosing a new initial value of the rate close to the average value of the minimum cost variance index.