The paper analyses a network with given input and output flows in each of its nodes. The basis of this analysis is the algorithm for determining the set of solutions of the linear equations system, using the Gaussian method. The power of the set determines the structural entropy of the system. By introducing uncertainty into the value of part of the information flows, the deviation of the network from its equilibrium state is simulated. The set of potential solutions, as a part of the total set of the system solutions, determines the statistical entropy of the system. The probability entropy is calculated for a network with four nodes and a total flow of 10 erlangs with a sampling step of 1 erlang. Calculated entropy values for 1, 2, 3, and 4 uncertain flows out of a total of 16 flows that are transmitted between nodes of the fully connected network. As a result of the conducted statistical analysis of entropy values, the optimal number of statistical intervals for entropy values is determined: 4, 11, 24, and 43 intervals for 1, 2, 3, and 4 uncertain flows, respectively. This makes it possible to highlight the set of flows in the system that have the greatest influence on the entropy value in the system. The obtained results are of practical importance, as they enable the detection of deviations of the network from its equilibrium state by monitoring the passage of traffic on individual branches of a complex telecommunication network. Since, as shown in our previous works, the task of determining the complete set of solutions of the system for the number of nodes greater than 4 has a significant computational complexity, the application of the algorithm to such networks requires an increase in the discretization step of the values of information flows in the network. Another way to reduce computational complexity can be to reduce the set of analysed solutions to a subset of solutions close to the equilibrium state of the system.
Read full abstract