Research On Intelligent Diagnosis Technology Of Low-overhead Network Fault Based On Machine Learning

Due to the high real-time nature of mobile communication networks,high throughput,high scenario complexity,easy coverage of fault valid data,and non-regular and non-linear communication information data,current mobile communication network fault diagnosis solutions often face problems such as high resource overhead,slow response,low accuracy rate and lack of flexibility.To address these problems,this paper provides a low-overhead,fast,accurate,lightweight and general intelligent fault diagnosis implementation from the perspective of fault data source acquisition and diagnosis strategy by using flow-limiting algorithm,SOM algorithm,K-Means algorithm,TF-IDF algorithm and 3GPP-related knowledge.The following overheads are mainly reduced: 1.fault judgment point overhead;2.fault information source collection overhead;3.fault reproduction cost overhead;4.overhead of mapping fault diagnosis results to official 3GPP interpretations.In addition,because this scheme complies with 3GPP protocol,the diagnosis results are more accurate,authoritative and convincing.Finally,we have verified and evaluated this scheme in a real communication environment.The experimental results show that this scheme greatly reduces the fault diagnosis overhead and significantly improves the correct rate and efficiency of fault diagnosis compared with the traditional scheme,and a large amount of experimental data proves the effectiveness and practicality of this scheme.