Research On Network Traffic Classification Technology Based On ELM

Big data is the product of this era.With the advent of the cloud era,many unstructured and semi-structured data require a framework like MapReduce to distribute work for hundreds of computers.The network application is diversified,and the surge of network traffic and the complexity of the network system bring severe challenges to network management.Network traffic classification is a technology that identifies traffic characteristics from data and analyzes protocols and applications.This technology is of great significance for maintaining network security,ensuring network operations,and improving network service quality.Network traffic classification generally has three steps: collecting traffic data,extracting traffic characteristics,and matching classified traffic.Due to the uncontrollable conditions,there will be messy attributes and linear correlation between attributes.Therefore,the content of this paper is mainly divided into two aspects: attributes reduction and the use of algorithms to match classified network traffic.The effect of PCA and rough set combined with extreme learning machine on network traffic classification is studied.The main work of the paper includes:1.Data preprocessing and the dimensionality reduction algorithm are studied.Data preprocessing integrates the data set and does the premise work for constructing the network traffic classifier.The problem of the redundancy attributes are solved by PCA and RS.2.The feedforward neural network algorithm(BP neural network)in network traffic classification is studied.A BP network traffic classifier is designed.The BP neural network has the ability to approximate arbitrary functions to classify network traffic data,but the parameter settings traffic classifier is complex and the training time cost of the BP algorithm is high.3.The application of the combination of extreme learning machine(ELM)and data reduction in network traffic classification is studied.A ELM traffic classifier which does not need to set the learning rate is designed.The ELM traffic classifier combined with data reduction improves the learning speed and classification accuracy of the classification.4.The hierarchical extreme learning machine(H-ELM)is studied as a traffic classifier to classify network traffic.The H-ELM traffic classifier itself has a feature extraction phase.The experimental results show that the designed H-ELM network traffic classifier greatly improves the accuracy of the network traffic classification.5.The online sequential extreme learning machine(OS-ELM)is studied as a traffic classifier to classify network traffic.OS-ELM has the ability of batch processing,which can process data in batches and train current data blocks in time,and then release the current space to meet the needs of online learning and improve the efficiency of network traffic classification.