Quality of Service Measurement and Prediction through AI Technology

With the development of Information Technology (IT) and Software-Defined Networking (SDN), Communications Service Providers (CSPs) can collect much information from telecommunication circuits. However, some of the existing circuit measurement has not been used well. For CSPs, it is important to find more efficient ways of utilizing the circuit measurement, e.g., delay, jitter, packet loss, and speed-test results for customer satisfaction. One of the most popular ways is to use speed-test tools (such as Speedtest online) to measure the service rate of the application layer. However, it is difficult to justify whether the telecommunication circuit is normal or not. For example, when the speed-test result of a specific circuit is 90 Mbps, the physical line rate may be 100Mbps. To address the above issues, we first investigate the measurement and management mechanisms of the existing telecommunications networks, including the core components and protocols. In this paper, we leverage artificial intelligence (AI) technologies to predict whether customers complain or not. We evaluate the proposed AI model by using the real data from telecommunication circuits and analyze the key performance metrics.