Modeling of self-similar network traffic using artificial neural networks

Self-similarity is a phenomenon which has come into computer networks literatures during last two decades and plays a significant role in modeling of computer network traffics. It is generally accepted that computer network traffics are self-similar and they are dissimilar to Poisson-based traffics. Computer network models exert a considerable influence on improving quality of service. Therefore, self-similarity should be considered in traffic models in order to acquire more appropriate QoS. In this paper, we propose a novel model for generating self-similar traffic. Our model includes a multi-layer perceptron neural network and a random error generator. This model has two phases: Firstly, the model is trained with real network traffic. Secondly, with the assistance of the random error generator, it generates traffic which is as self-similar as the real traffic. The implementation and the results validate this model through drawing a comparison between the Hurst parameter of the generated traffic and the real traffic.