Research And Application Of Key Issues In Campus Network Behavior Mining Under Big Data Environment

The Internet has an important influence on People's Daily life,as the way people use the Internet increasingly diversified methods and the application type,the user's network behavior of the resulting data quantity is also growing rapidly.Dig out the available information from the user network behavior data,not only has great research value and commercial value,also can provide decision support for the government and provide guidance for people's production and life.The contents of this paper mainly focuses on the network behavior data acquisition,modeling,analysis and mining analysis and research are carried out in the big data environment,combined with the course grade of the students of data mining,analyzes the influence of network behavior on learning achievement.First of all,Aiming at the problem that the performance of the network data collected by the general server is not enough,this paper analyzes the two aspects of hardware and software,a comparative test of the new packet capture scheme is made,a method of collecting data packets by line-speed using a universal server in a Ten-gigabit network environment is presented,and the protocol identification method is survey.Secondly,in the aspect of descriptive analysis,in order to solve the problem that the network behavior data is difficult to analyze in the big data environment,a multidimensional analysis model is designed,the Pentaho platform is used to clean and transform the data,a visual observation environment of WEB is implemented which meets the need of descriptive analysis in the research work.Thirdly,In attribute value reduction and data preprocessing,according to the traditional treatment method for massive network behavior data exists the problem of low efficiency,parallel operation of network behavior data using Pentaho platform components called Hadoop and R plugin to reduction of network behavior data by particle size amplification method and box plot method to solve the processing efficiency the problem of low data processing.Finally,in the mining network behavior influence on student's achievements,three score prediction models are established by using Linear SVM,GBT and KNN,respectively.Then these models are tested by the final exam results.After the network behavior as the factors that affect academic performance were introduced,through the use of decision tree algorithm,this paper finds out the network behavior influence on course grades and GPA,and finds out some network applications which have a great influence on learning,as well as the threshold which need for students to be intervened.