New Algorithm for Predicting the States of a Computer Network Using Multivalued Dependencies
Oleg I. Sheluhin, Andrey V. Osin, Dmitry I. Rakovsky · Automatic Control and Computer Sciences · 2023
Abstract The algorithm that allows based on the analysis of monitoring historical data to predict the state of the network has been proposed. It is aimed to evaluate the state of the network using the service level indicators Service Level Objectives, SLO. The algorithm is based on the search for statistically stable multivalued dependencies. For forecasting, combinations of arguments are calculated, obtained by imposing masks, on the basis of which, in turn, statistical patterns are formed. To search for repeating patterns a history data size is determined. Each stage of the algorithm is illustrated with pseudocode.