Alarm Correlation Analysis Base On Business

In recent years, with the gigantic development of modern telecommunication technology, the telecommunications industry, an irreplaceable part of national economy, has made substantial achievements. Network optimization and network fault management are vital means to ensure the quality of network. The traditional network optimization mainly depends on of engineers’ experience to handle it and draw lessons. With telecommunication network expanding in tremendous size, managing network failures, becoming far tougher, always costs enormous money every year. Besides, disadvantages of the traditional pattern have gradually begun to hinder network management or led to ineffective conditions.In this paper, we conduct a research on the communication network alarm correlation based on data mining, elaborate a method for preprocessing original alarm data, and propose a clustering algorithm based on density, which divides data sets into transaction sets. Through the analysis of association rules, we conduct data mining on alarm data from Chinese United Telecommunications Corporation and design a specialized platform to excavate alarm data based on association rules.The main contents in this paper are as follows:Firstly, we analyze alarm correlation and alarm data mining methods, elaborate several correlation algorithms and compare the pros and cons of these algorithms. Then, we preprocess original alarm data, consisting of two parts:firstly, we summarize characteristics of alarm data after analyzing regular pattern of alarm data and accordingly design a multi-filter model aiming to filter original alarm data to improve data quality and mining efficiency, and to reduce data amount; And then, we design an improved DBS CAN (Density-Based Spatial Clustering of Applications with Noise) algorithm to convert data to a transactional data set, and verify the validity of the improved algorithm.Finally, we design a specialized platform to excavate alarm data through association rules, and propose a concrete execution program for the platform.