A network diagnostics method based on pattern recognition algorithms

Abstract

This report deals with the problem of designing and building of a computer network diagnostic system. Diagnostic content problems are reviewed, as well as ways of their solution based on mathematical and computer modelling methods. A traffic analysis-based diagnostic method is suggested for process statuses in a computer network. The method is based on algorithms of the mathematical pattern recognition theory. To build a diagnostic system, a multi-level model building and verification arrangement is suggested. Methods and techniques for data collection and preliminary analysis are considered.