Research of Intelligent Data Analysis Methods for Identification of Computer System State

Classification is one of the methods of the intelligent data analysis (data mining), which allows to study the existing data set and create a model for each class based on its data features. To increase the accuracy of classification, ensemble methods are used, including some of the most effective ensemble methods based on boosting and bagging. This study presents the results of a comparative analysis of machine learning methods for classifying the state of a computer system. The ensemble methods based on boosting and bagging were modified using a special procedure for selecting the main classification parameters: the number of trees (the number of learners), the number of branches (the number of splits or leaves nodes), and the learning speed (the learning rate). Modified methods were tested with the following initial data of the computer system functioning: loading of the central processor and memory, the amount of incoming and outgoing traffic (network utilization), the number of read / write operations to disk, the statistical data of the system events analysis (the number of operations with the system registry, file system, number of processes, etc.). It was found that the classification accuracy with the test sample was more than 99%.