Automatic Fault Detection with Bayes Method in University Campus Network

Abstract

In recent years, the use of the internet has become widespread with developing technologies. Internet is used for many needs, especially social media. Today, internet is needed for remote use of electronic devices used in homes and offices. Continuous access to the internet is very important for the quality of life of people. In this study, a proposal was made for early detection of basic faults that may occur in the lines of internet access devices. The recommended method was carried out in the university campus environment. Fiber cable traffic, which provides internet streaming between departments within the campus, is constantly monitored via switch. As a result, a set of data has been received from the switch ports to which the fiber cables are connected. Failure detection has been performed taking into account the changes occurring in this data. Eliminating the malfunctions that may occur in the Internet line requires long time and workload. Thanks to this developed method, maintenance studies are made by providing early detection of failures that can occur in university campus environment. A software has been developed to monitor the uplinks of the switches in the system rooms and retrieve the data. The Bayesian classifier has been used to process the results and obtain the results. Over 90% classification success was achieved using the Naive Bayes method.