COMPARATIVE ANALYSIS OF MACHINE LEARNING METHODS FOR NETWORK TRAFFIC BINARY CLASSIFICATION

This research offers a a comparative analysis of different machine learning methods used to solve the problem of dividing network traffic into two separate categories. The study considered such algorithms as logistic regression, support vector machine, decision tree, random forest, gradient boosting, single-layer and multilayer perceptrons. Experiments were performed on the CIC-DDoS2019 dataset, which contains more than 75 characteristics of normal and abnormal traffic. Metrics of accuracy, recall and F-measure were used to evaluate the models. This made it possible to get a comprehensive understanding of the classification quality and compare the effectiveness of each algorithm. The project was implemented in the Jupyter Notebook environment, using the Python3 programming language and its libraries for data preprocessing, visualization and implementation of machine learning algorithms. The accuracy of anomaly detection was tested using various supervised learning models.