Predicting the Network Traffic flow using Long Short-Term Memory (LSTM)

Abstract

With the information pace of fast connections expanding, the effect of nonlinear variables turns out to be increasingly noticeable, which bring incredible sign uprightness difficulties to the examination and plan of rapid connections. Notwithstanding, there are two holes in the investigation of the nonlinear practices of the rapid connections. One is that there is no recreation stage which can deftly manage various types of nonlinear practices. An elective fast connection demonstrating and examination technique is recommended dependent on the framework ID approach, rather than utilizing industry standard models. This methodology is particularly advantageous for the framework structure/investigation in that the means are basic and the philosophy is entrenched in other control. The other is that there is no metric which can measure the degree of the nonlinearity of the connections. Focusing on these two issues, this paper first models a nonlinear rapid connection dependent on Simulink, which can deftly change the nonlinear practices and investigate the effect of nonlinear practices on the presentation of the fast connections. At that point a measurement is exhibited base on the chief segment examination (PCA) to evaluate the degree of the nonlinearity of the fast connections. Re-enactment results show that the measurement can viably catch the nonlinear seriousness of the connections.