Research on Deep Learning and Other Methods Based on Abnormal Traffic Detection in Complex Network Environment

Abstract

With the increasingly complex structure of complex network environments, how to use intelligent methods to diagnose network faults has attracted more and more attention from operators. Self-organizing network (SON) self-configuration, self-optimization and self-healing functions can be used to improve the self-organization capability of wireless networks, replace the manual intervention of high-cost network operators, and realize automatic anomaly detection and diagnosis, thereby effectively reducing the network Deployment and operating costs. In this paper, based on multivariate time series analysis and self-organizing neural network deep learning method, an abnormal traffic prediction and abnormal detection method for complex network environment is proposed. This method uses wireless channel quality indicator (CQI) to determine the traffic performance status and can respond to Reasonable system changes (concept drift). It can be seen from the application results in the actual operating LTE network that this method can effectively predict the traffic behavior trend and detect the irregular traffic behavior from a large amount of data in a self-organizing manner.