Application Of Data Mining Technology In University's Network Operation And Maintenance Management System

Network Operation and Maintenance, is a field packed with sizable hidden knowledge & expertise. Besides the knowledge already standardized by network experts and scholars, there exists many unknown, or undiscovered know-how and methods. Whereas the present Network Operation and Maintenance System, without induction and further mining for the service accumulated data, cannot lead the way to unveil hidden know-how and methods behind the network malfunction information. Lacking the capability of comprehensive analyzing and decision-aiding, the system cannot actively facilitate improvement of network service. The combination of data mining technology with Operation and Maintenance System, can satisfy the above needs.The past studies mainly utilized the network system's error message & performance information for data mining, and focused on discovering the hidden connections between these info, so as to locate the malfunction source. However, in the network systems of universities, as there are relatively few numbers of devices, the data size is not big enough for analyzing, while the network failures caused other than equipment factors were on the rise. Therefore, this type of data mining cannot suffice needs.This study carried out in-depth research and study of data mining related theories and technology, and in accordance with the inadequacy of present Network Operation and Maintenance System, a network malfunction analyzing and predicting module was designed, mining the actual malfunction data accumulated in University Network Operation and Maintenance System. Based on the characteristics of network malfunction and algorithm needs of data mining, the suitable source data were extracted from the Network Operation and Maintenance System. During the data warehouse building process, computer network and data warehouse related theories & technologies were applied in combination with the actual operation experience of Network Operation and Maintenance System. Therefore completing the ETL process such as source data cleaning and transforming, and satisfied the data structure requirements of data mining algorithm. With the cross-industry standard process for data mining, three algorithms, namely association rule, time series, decision tree, were respectively employed for induction and in-depth mining of network malfunction data warehouse. Through correlation analysis, the potential relationship and interaction between network malfunctions were studied; through decision tree, network malfunctions were classified and general rules induced; through time series study, statistics were analyzed, network malfunction trend predicted and correlated analysis model built. Finally, comparing and analyzing results predicted by malfunction analysis & prediction model with the actual service data, analysis and prediction of three aspects concerning network malfunctions were realized, namely the network malfunction diagnosis, trend of happening prediction, potential impact study, which was impossible for the present system.After testing, it is found out that the knowledge and methods found through network malfunction data mining, would be helpful for network management in malfunction diagnosis, prevention of malfunction happening, as well as allocating and relocating network resources. In this way the information advantages of Network Operation and Maintenance System can be bring into full play, and enable the system to display a more active role to enhance network service quality.