Study of Methods for Improving the Meta-Algorithm of the Bagging Classifier

In this paper, we explored approaches that improve the performance of ensemble bagging classifiers for identifying the state of a computer system. The following algorithms are considered: Ensemble pruning, Advanced Voting Algorithms, Dynamic Voting Strategies, Confidence Calibration, Adaptation through Meta-Features and Meta-Learning, which allow improving the voting procedure of the bagging meta-algorithm. Artificial data was generated as initial data, which complicates the classification task and contains an increased amount of noise. In the Google Collab environment, software models of algorithms have been developed, and their quality has been assessed. It was found that the use of the Ensemble pruning and Adaptation through Meta-Features algorithm is the most qualitative. In addition, the Ensemble pruning algorithm reduces the number of basic ensemble classifiers, and, as a result, increases the efficiency of computer system identification. Based on the results of the study, a method for identifying a computer system was proposed through the integrated use of a bagging classifier and an optimization procedure based on the Ensemble pruning algorithm.