A wireless network fault detection method based on self-learning

Abstract

The invention discloses a wireless network fault detection method based on self-learning, which is specifically as follows: collect network key performance index KPI records and save them as a data set; use SMOTEENN method to unbalance the data set to obtain balanced Balance the data set; learn the base vector from the unlabeled auxiliary data set through the sparse autoencoder; express the balanced balanced data set as a linear combination of the base vectors, and pass the support vector on the balanced data set under this new representation The machine SVM method is used to train the classification model of normal and fault categories; the established classification model is used to classify the KPI data records generated in real time by the network, and then the purpose of fault detection is achieved. The invention detects network faults more accurately and effectively; and the form of self-learning is convenient for migration, and a fault detection model can be quickly obtained in a new network environment, thereby improving the fault detection efficiency of previous methods.

Description

technical field [0001] The invention relates to the field of network technology in wireless communication, in particular to a wireless network fault detection method based on self-learning. Background technique [0002] Self-healing is one of the important functions of wireless self-organizing networks, and accurate and effective fault detection is an important step to realize the self-healing function. In the current wireless communication network, the figure 1 The macro base station-HNB heterogeneous structure is used to improve the indoor coverage of the macro base station and provide capacity gain. Compared with traditional homogeneous networks, femtocells deployed by users may be more prone to configuration problems such as improper transmission power settings and channel conflicts due to operational errors; in addition, the number of femtocells is far greater than that of macrocells, and the distributed architecture also makes it easier for families to malfunction. ...

AI Technical Summary

Problems solved by technology

[0003] The current fault detection methods for homogeneous networks are often not suitable for heterogeneous networks. The main reason is that fault detection in homogeneous networks is mostly based on classification methods, and the accuracy of classification depends to a certain extent on the amount of historical network data.

However, the available data of femtocells in heterogeneous networks is often limited

There are two reasons for this. On the one hand, femtocells deployed indoors usually serve fewer users and obtain less data than homogeneous networks; on the other hand, the deployment of femtocells is determined by users. Switching, redeployment and other operations lead to dynamic changes in the network topology, and the historical data accumulated by the base station for a long time cannot be directly used to analyze the current status of the base station, that is, the timeliness of the historical network data is low

To sum up, in the prior art, there are problems such as scarcity of user data of femtocells in heterogeneous networks and easy outdation of historical data

Images