A performance evaluation of general queueing systems by machine learning

This paper presents a method for performance evaluation of general queueing systems that is important for designing IoT (Internet of Things) data processing systems. IoT has been paid great attention all over the world. There are a lot of IoT devices that connect Internet. In addition, IoT devices are many kinds. Hence, IoT data processing systems have to handle massive, many kinds of data. From the above, when we design IoT data processing systems, it is important to evaluate performance of general queueing systems. However, in general queueing systems, the exact solution that can evaluate performance is not available. Alternatively, we can evaluate it with discrete simulation. However, it spends much time. From the above reasons, we evaluate performance of general queueing systems by machine learning instead of discrete simulation. In addition, we validate what kind of teacher data we should use for machine learning. © 2019 The Institute of Electrical Engineers of Japan.