Multi-method Fusion Intelligent Terminal Detection And Application Identification

With the continuous development of Internet,mobile terminal devices have gradually become essentials for people,and they are inseparable from each other.In order to better understand the information of mobile users,mobile terminal device model identification technology was designed.At the same time,with the increase of network users,the increase of network traffic and the increasing number of network applications,network traffic classification technology has gradually developed.The network traffic classification technology can identify the application categories represented by the data flows,thereby optimizing the network resources and improving the network service quality.In this regard,did the following work:Aiming at the problem of low recognition rate of terminal models,a terminal device identification technology based on Jaccard similarity coefficient is proposed.Before the identification phase,the design combines the multi-granularity UA word segmentation method and the TF-IDF algorithm filtering technology to complete the extraction of the key sentences including the terminal device model.Finally,the Jaccard similarity coefficient is introduced into the terminal identification to obtain the terminal device model information.Compared with WURFL and based on pattern detection and recognition technology,the recognition rate of the device has been improved,and at the same time,it has achieved certain effects against the changing operating environment.For the classification and identification of the network data generated by the terminal equipment in the network,in terms of feature processing,most algorithms only consider the relationship between features and classification,or combine the classifier to perform feature processing.These methods cannot be largely remove redundant features and characteristics are not relevant for classification.In this regard,a feature weighted clustering feature selection algorithm based on information gain rate weight correlation coefficient is proposed to improve the classification efficiency of network traffic.Firstly,considering the relationship between features and features,a metric based on information gain rate weight correlation coefficient is designed.Then,using the feature selection algorithm proposed in this thesis,the number of redundant and unrelated features is greatly reduced.Then,using the evaluation results of the SVM classifier,the final feature subset is obtained.Finally,the simulation results show that the featureselection algorithm proposed in this thesis performs well,and its average f-score value is 99.30%,which is improved compared with CFS,Relief-F and selected contrast experiments,and the feature selection algorithm proposed in this thesis.It performs better on the J48 and NB classifiers.