The object of the study is the process of identifying the state of a computer network. The subject of the study are the methods of identifying the state of computer networks. The purpose of the paper is to improve the efficacy of intrusion detection in computer networks by developing a method based on transformer models. The results obtained. The work analyzes traditional machine learning algorithms, deep learning methods and considers the advantages of using transformer models. A method for detecting intrusions in computer networks is proposed. This method differs from known approaches by utilizing the Vision Transformer for Small-size Datasets (ViTSD) deep learning algorithm. The method incorporates procedures to reduce the correlation of input data and transform data into a specific format required for model operations. The developed methods are implemented using Python and the GOOGLE COLAB cloud service with Jupyter Notebook. Conclusions. Experiments confirmed the efficiency of the proposed method. The use of the developed method based on the ViTSD algorithm and the data preprocessing procedure increases the model's accuracy to 98.7%. This makes it possible to recommend it for practical use, in order to improve the accuracy of identifying the state of a computer system.

This paper examines traditional machine learning algorithms, neural networks, and the benefits of utilizing ensemble models. Data preprocessing methods for improving the quality of classification models are considered. To balance the classes, Undersampling, Oversampling, and their combination (Over + Undersampling) algorithms are explored. A procedure for reducing feature correlation is proposed. Classification models based on meta-algorithms such as SVM, KNN Naive Bayes, Perceptron, Bagging, Random Forest, AdaBoost, and Gradient Boosting have been thoroughly investigated. The settings of the base classifiers and meta-algorithm parameters have been optimized. The best result was obtained by using an ensemble classifier based on the Random Forest algorithm. Thus, an intrusion detection method based on the preprocessing of highly correlated and imbalanced data has been proposed. The scientific novelty of the obtained results lies in the integrated use of the developed procedure for reducing feature correlation, the application of the SMOTEENN data balancing method, the selection of an appropriate classifier, and the fine tuning of its parameters. The integration of these procedures and methods resulted in a higher F1 score, reduced training time, and faster recognition speed for the model. This allows us to recommend this method for practical use to improve the quality of network intrusion detection.

The object of the study is the process of identifying anomalies in the operation of a computer system (CS). The subject of the study is ensemble methods for identifying the state of the CS. The goal of the study is to improve the performance of ensemble classifiers based on heterogeneous models. Methods used: machine learning methods, homogeneous and heterogeneous ensemble classifiers, Pasting and Bootstrapping technologies. Results obtained: a comparative analysis of the use of homogeneous and heterogeneous bagging ensembles in data classification problems was carried out. The effectiveness of various approaches to the selection of base ensemble classifiers has been studied. A method for identifying the state of a computer system, based on the heterogeneous bagging ensemble was proposed. Experimental studies made it possible to confirm the main theoretical assumptions, as well as evaluate the efficiency of the constructed heterogeneous ensembles. Conclusions. Based on the results of the study, the method for constructing a heterogeneous bagging ensemble classifier, which differs from known methods in the procedure for selecting base models was proposed. It made possible to increase the classification accuracy. Further development of this research could include the creating and integration of dissimilarity metrics as well as other quantitative metrics for a more accurate and balanced base model selection procedure, which would further improve the performance of the computer system state classifier.

The object of the study is the process of classifying objects in scientific problems. The subject of the study is methods aimed at assessing the effectiveness of multiclass classification. The goal of the study is to study the classification process and develop a classifier evaluation module to increase the speed of such evaluation and reduce the time to build complex machine learning classifiers. Methods used: methods for evaluating machine learning classifiers, methods for constructing ROC curves, principles of parallel and distributed computing. Results obtained: an analytical review of the scope of application of the classification quality assessment module in the field of humanities, technical and economic sciences was conducted. Existing classification quality assessment metrics were considered and mathematical descriptions of metrics were formed for the multi-class case. Software was developed that implements the proposed mathematical descriptions using parallel calculations and optimization of identical operations. The developed module was tested for reliability. Conclusions. According to the results of the study, methods for effective classification quality assessment is proposed, which allows reducing the time for assessing the quality of multi-class classifiers by 40% compared to the classical methods. The development of this module opens up broad prospects for further research in the direction of improving the quality of classification, which will contribute to the development of various spheres of human activity and increase the efficiency of solving tasks related to data analysis.

Предметом статті є дослідження методів ідентифікації стану комп’ютерної системи. Метою статті є розробка методу класифікації аномального стану комп’ютерної системи на основі ансамблевих методів. Завдання: дослідити використання алгоритмів побудови дерев рішень: REPTree, Random Tree, J48, HoeffdingTree, DecisionStump та ансамблів дерев рішень на основі беггінгу та бустінгу для ідентифікації аномального стану комп’ютерної системи на основі аналізу подій операційної системи. Використовуваними методами є: методи штучного інтелекту, машинного навчання та ансамблеві методи класифікації. Отримано такі результати: досліджено методи ідентифікації аномального стану комп’ютерних систем на базі ансамблевих методів, а саме беггінгу, бустінгу та класифікаторів: REPTree, Random Tree, J48, HoeffdingTree, DecisionStump для ідентифікації аномального стану комп’ютерної системи. Розроблено набір різних класифікаторів та ансамблів класифікаторів, виконано їх навчання та кросвалідацію. Виконано оцінку ефективності розроблених класифікаторів. За результатами досліджень запропоновано ансамблевий метод класифікації стану комп’ютерної системи на основі алгоритму побудови дерева рішень J48. Висновки. Наукова новизна отриманих результатів полягає в створенні ансамблевого методу класифікації стану комп’ютерної системи на основі дерева рішень, що надає можливість підвищити надійність та швидкість класифікації.

The object of the study is the process recognition of pedestrian red light traffic. The subject of the study are the methods of process recognition of pedestrian red light traffic. The purpose of the paper is to improve the efficiency of real-time pedestrian red light traffic recognition model. The results obtained. The pedestrian red light traffic recognition model based on real-time object detection architecture YOLOv8 was proposed. The architecture and characteristics of YOLOv8 model, including its improved network structure, multi-scale detection ability, and adaptive anchor adjustment were introduced in detail. To demonstrate the efficiency and benefits of applying the YOLOv8 model, its performance was evaluated in various scenarios. Conclusions. Experiments have confirmed the efficiency of the proposed method. The use of the developed method based on the YOLOv8 architecture allowed to increase precision up to 0.935. Overall, the average performance across all categories is 0.851, which means that the model has a relatively high detection accuracy. In addition, model has a high-speed index.