METHOD OF CONTEXT-AWARE CONTROL OF THE ADAPTATION PROCEDURE FOR AN INTELLIGENT COMPUTER SYSTEM

The article proposes a method of context-aware control of the adaptation procedure for an intelligent computer system, which simultaneously takes into account the magnitude of contextual changes, the relevance of these changes to the intelligent system’s current task, the system’s sensitivity to contextual changes, the expected consequences of canceling the execution of the adaptation procedure, the mismatch between the intelligent system’s current configuration and the context, and the cost of performing the adaptation procedure. An integrated assessment of the feasibility of performing the adaptation procedure and a corresponding decision-making mechanism are also proposed, taking into account the current intensity of contextual changes, the computer system’s load level, and the current performance level of the intelligent computer system. The results of the experimental study show that the proposed method increases the informativeness and accuracy of decisions regarding the execution of the adaptation procedure; specifically, in a test scenario with a large change in context of low relevance, it demonstrated the ability to account for both the low relevance of the context and the low sensitivity of the intelligent system to contextual changes, thereby enabling the proposed method to reduce the average proportion of decisions to perform the adaptation procedure to 8.83%, ensuring significantly better decision accuracy compared to the baseline method.