NETWORK ASSISTANT BASED ON ARTIFICIAL INTELLIGENCE Russian patent published in 2021 - IPC H04W24/02 H04W24/04 G06F15/173 

FIELD: machine learning.

SUBSTANCE: invention relates to a machine-readable bearer and a method for analyzing data to eliminate network failures. The method includes receiving on a data adapter platform implemented on one or more computing nodes performance data related to user devices and network components of the wireless operator network from several data sources, wherein performance data includes one or more from the following: data about network component performance, data about user device performance, social media data, alarm signal data, data about failure elimination requests or data of key performance indicator; identifying in an application for searching and network failure eliminating performed on one or more computing nodes at least one quality of service (hereinafter – QoS) problem affecting the network performance in one or more geographical region with low performance, wherein this QoS problem negatively affects the performance of user devices or network components so that the performance level drops below the pre-set threshold; analyzing using the application for searching and network failure eliminating performed on one or more computing nodes performance data using trainable model of machine learning for determining the predictable main cause of QoS problem of the network affecting the network performance, wherein the trainable model of machine learning uses several types of machine learning algorithms for analyzing performance data; providing through the application for searching and network failure eliminating performed on one or more computing nodes predictable main causes for providing through the user interface; refining the machine learning model based on feedback from the user concerning the accuracy of the predictable main cause, wherein the refinement includes re-training the machine learning model based on at least one of the following: the training unit, which is modified based on the mentioned feedback, or one or more rules of selection of a modified algorithm; providing priorities of eliminating network failures for each of QoS problems of the network to implement a decision, which eliminates the predictable main cause of QoS problem of the network; and refining the machine learning model based on feedback from the user relatively to the accuracy of priorities of eliminating network failures, wherein the refinement includes re-training the machine learning model based on at least one of the following: the mentioned training unit, which is modified based on the mentioned feedback, or one or more rules of selection of the modified algorithm.

EFFECT: increased accuracy of network failure prediction.

10 cl, 6 dwg