Purpose. Study on the process of combining several methods for determining the quality indices of human-machine interaction, containing various configurations for determining the weight coefficients in an ensemble. Methodology. The process of diagnosing the quality of the interaction of a human-machine system with four elements of subsystems is studied using the example of the system “Operator–Machining Center – Control Program – Safe Environment”. The main hypothesis of the study is the combination of several methods for determining the quality indices of human-machine interaction, containing different configurations for determining the weight coefficients in the ensemble. A combined method for diagnosing the quality of interaction between human-machine systems based on ensemble models, which include non-ensemble ones, has been proposed. The ensemble index has been determined by averaging the non-ensemble indices. The defined ensemble indices and element scores of the four subsystems are used as input scores to a multiple regression model to generate prediction. Findings. Four combinations of ensemble indices have been developed and implemented in software, which are characterized by a minimum value of the standard deviation compared to the existing ones. According to the results of experimental verification, the proposed models demonstrate the value of the standard deviation of 0.1404; 0.1401; 0.1411; 0.1397, and the existing ones are 0.1532; 0.1535; 0.1532; 0.1532. Originality. The combined ensemble method for diagnosing the quality of interaction between elements of subsystems takes into account linear models with non-linear variables and different ways of determining weight coefficients. Practical value. The scenario for the practical use of the results obtained is a possible option for optimizing production, where, depending on the final result, specialists can adjust the value of a particular subsystem to achieve the desired result.