METHOD TO CONTROL COMPUTER NETWORK Russian patent published in 2015 - IPC G06F15/177 G06F11/30 G06N3/02 

FIELD: information technologies.

SUBSTANCE: predictive neuron networks are formed on the basis of parameter values received as a result of polling working objects, which determine their condition; the prediction of values change in a certain fixed interval of time is formed. Memorising neuron networks are formed, which serve as functions of an indicator of necessity for retraining of the appropriate predictive neuron network, if values produced as a result of monitoring considerably differ from values used in the training sample. After retraining the predictive neuron networks may form prediction of parameter changes with the specified accuracy level, on the basis of which control actions at objects of the computing network are performed.

EFFECT: reduced time of outage of working objects of a computing network due to increased accuracy of prediction of main parameters change.

9 cl, 7 dwg