Proposing a Hybrid Data Mining Approach for Network Traffic Analysis

N etwork traffic is an important parameter for evaluating the running status of a network. Network data is a nonlinear time series with time variable characteristics, long-term correlation, self-similarity, suddenness, and chaos. Network traffic forecasting is a reliable way to secure communications in a network management and monitoring system. The prediction process is based on network traffic features analysis in the past and present, then the creation of internal structure rules, and then the building of a model for predicting the characteristics and trends of future traffic. Recently, data mining methods have been used effectively to predict traffic in the network. In this research, network traffic prediction using four methods in data mining has been presented. Neural Networks, simple Bayesian, Boosting, and Bagging. According to the results of this study and comparing the methods with each other, the better evaluation criteria for predicting network traffic is the Bayesian method.