Reinforcement machine learning model for sports infrastructure development planning

The paper considers the actual task of planning the rational development of sports infrastructure in conditions of limited resources. The development of a mathematical model for the evaluation of sports infrastructure projects and the schedule for their implementation was carried out. To evaluate projects, it is proposed to use methods of multi-criteria decision analysis based on fuzzy preference areas. The search for the optimal parameters of the proposed model is difficult due to the presence of binary variables that make the problem NP-hard. To find a solution close to the optimal one, a machine learning model with reinforcement is proposed. Software has been developed that allows both ranking projects and determining the schedule for their implementation, taking into account available resources and needs. An algorithmic and software solution based on a machine learning model with reinforcement is invariant with respect to the subject area and can also be used in other combinatorial optimization problems. On the example of the problem of choosing regions for the construction of basketball courts, computational experiments were carried out for the proposed solution.