Proceedings of 8 th IOE Graduate Conference Peer Reviewed ISSN: 2350-8914 (Online), 2350-8906 (Print) Year: 2020 Month: June Volume: 8 Network Device Status Detection using ANFIS-based Classification for ISP Networks Upgrade Planning Prabha Shastri a , Babu Ram Dawadi b , Shashidhar Ram Joshi c a, b, c Department of Electronics and Computer Engineering, IOE, TU, Nepal Corresponding Email: a 074msice012.prabha@pcampus.edu.np, b baburd@ioe.edu.np, c srjoshi@ioe.edu.np Abstract With recent advances in technology, a lot of network equipment is always being brought and replaced as per needs. So the management of the networking equipment is a significant challenge in terms of operational cost and manpower. Many software are available to monitor the network health, but for the replacement of networking equipment, there is no mechanism developed till now. In this study, an adaptive neuro-fuzzy inference system was developed (ANFIS), based on input parameters such as based on input parameters like throughput, CPU usage, memory, end of life [EoL], end of support [EoS], and energy consumption to predict replacement of network devices like router and switch. This research will help internet service providers to decide whether to replace or upgrade equipment to ensure proper network infrastructure is functional. The simulation result shows that ANFIS gives a better result in terms of root mean square error. In case of router best ANFIS model was with gaussian membership function (gauss MF) having input membership function 2*2*2*2*2 and output membership function constant with testing RMSE of 0.1288. In case of switch best ANFIS model was with gaussian membership function (gauss MF) having input membership function 2*2*2*2*2*2 with output membership function constant with testing RMSE of 0.23663. ANFIS approach is very much useful and easy to implement. The evaluation showed the superiority of the proposed expert-based approach using ANFIS with hybrid learning for the prediction of the replacement of networking equipment. Keywords ANFIS, ISP network, router, switch, replacement planning Information is the building block for smooth communication. The company’s productivity depends on its capability to deliver useful information through proper communication. The systems and infrastructure used to ensure the proper communication requires constant maintenance to reduce the downtime and make the system running and operational. Small business and ISP owners spend a lot of money annually in devices to make networking systems operate efficiently, and then forget about them before something goes wrong. With time as the device starts to age, the performance starts degrading, and there might be a breakdown. In this situation, replacement of the hardware is the best possible solution instead of investment in repair. It is important to predict how long does the hardware of networking systems lasts. Managing thousands of networking devices is time-consuming and costly. If devices are not properly managed, then this leads to their rapid obsolescence. Without creating any replacement plan, if we replace old equipment, this might lead to complications [1]. We might face several critical issues with the compatibility of the newer device with the existing system. Newly purchased equipment requires extra cost for purchase, and some equipment might require extra cost for staff training. Good maintenance management will substantially reduce operational costs. Hence, it is imperative to provide a methodology to prioritize all networking equipment for replacement. In this paper, we propose the network device replacement plan to identify the parameters to develop and establish a knowledge-based reasoning engine and to model the device status detection system based on a fuzzy inference system, and to provide the recommendations based on the decisions obtained from the inference engine. ANFIS is used for classification [2] for device replacement planning. While making device status prediction regarding upgrade or replace the proposed approach should Pages: 1041 – 1048