Explainable Machine Learning in Service Management of Transport Corporation

Abstract

An effective tool for the formation of intelligent transport systems is machine learning (ML). But applying increasingly sophisticated machine learning technologies is creative. This allows us to speak about the significant subjectivity of the conclusions of ML. Thus, a gap is formed with the theory and practice of management based on ML. Intelligent technologies for managing transport systems using understandable machine learning are called upon to fill this gap. The article discusses an approach to the development of such a technology for a three-level service management system in a corporation. At its top level is the Boss, at the middle level is the Curator, at the bottom level is the Manager. The Boss should increase the scope of services provided by the corporation. But the Curator knows his abilities better than the Boss. In turn, the Manager knows his potential better than the Curator. Thus, both the Curator and the Manager can manipulate the scope of services they provide in order to get more incentives. To avoid this, a service management system is proposed that includes two explainable ML procedures. Sufficient optimality conditions for this control system are found. In their implementation, both the Curator and the Manager are interested in maximizing the scope of services provided. Such a management system provides algorithmic accountability, responsibility, trust and recognition of ML in the corporate team. The proposed approach is illustrated by the example of service maintenance repair of rolling stock.