Fault diagnosis method and system based on active learning self-organizing cellular network

Abstract

The invention provides a fault diagnosis method and system based on an active learning self-organizing cellular network. The method comprises the following steps: constructing a system simulation platform; simulating a network fault by using the system simulation platform to obtain a plurality of key performance index KPI data; selecting a first set number of KPI data from the plurality of piecesof KPI data as a training set, and taking the remaining KPI data as a test set; determining a learning engine according to the training set; and performing verification by using each piece of KPI datain the test set to obtain an output learning engine, so that the KPI data to be tested is subsequently input into the learning engine for fault diagnosis. According to the invention, the number of KPI data with fault causes required by fault prediction is reduced, and the accuracy of fault prediction is improved.

Description

technical field [0001] The invention relates to the technical field of fault diagnosis, in particular to a fault diagnosis method and system based on active learning self-organizing cellular network. Background technique [0002] In the past decade or so, the number of smart terminal devices has increased dramatically. In order to improve network capacity, coverage and service quality, operators have deployed a large number of small base stations. Here comes the great challenge. In order to reduce the operator's capital expenditure and operating expenditure, 3GPP has introduced an ad hoc network. The concept of the self-organizing network is to allow the cellular network to be self-configured, self-optimized, and self-healing. The fault diagnosis of the cellular network is an important part of the network self-healing. [0003] At present, machine learning is mainly used for cellular network fault diagnosis. A classifier model is trained through the data set formed by the ...

AI Technical Summary

Problems solved by technology

However, there are very few KPI data with fault causes in real life, because fault diagnosis experts basically do not write the specific fault cause into the KPI data recorded by the base station after completing the fault diagnosis, so there are very few KPI data with fault causes

In addition, since the base station is working normally in most cases, most of the recorded KPI data are fault-free, so this dataset is category-imbalanced, and a category-imbalanced dataset is directly used to train the classification In the prediction stage, the classifier will be biased towards the class that is predicted to account for a large proportion, so there is a problem of low accuracy of fault diagnosis

Images