Design Of Web User Behavior Analysis System Based On Distributed Structure

The mobile terminal application allows the size of the network users expand rapidly. Frequently access behavior has accumulated vast amounts of information which implies the lots of effective data. Due to differences in individual network users, Single user behavior does not constitute characteristics law. But when considering the customer groups, implicit feature will revealed, If anybody can accurately grasp the characteristics of the user community, and then divide them,so internet applications and service providers can provide personalized service and high value-added business for different custom groups in order to maximize the benefits network customer groups and network of service providers.This topic has designed a high-performance distributed network user behavior analysis system to divide user groupsFirst of all, Get Web content by using TFIDF word segmentation technology and then transforms them into web vector. Meanwhile obtain the context information through the WEB server which users can access. Eliminate redundancy through data preprocessing module, forming the only one data source with low redundancy.Second, Research and improve clustering methods in data mining techniques detailed, And implemented algorithm parallelization in Hadoop distributed processing framework we called it MapReduce In order to make it more suitable for mass data processing. Validate the MapReduce Performance improvement in parallel processing. The last, Design a framework for distributed user behavior analysis system, including data acquisition module, data preprocessing module, and the text clustering module. We show every module and the result achieves the major function of each module. Under the existing system performance test indicators we tested and evaluated this system. Above of all, I summarize the characteristics and shortcomings of the paper and proposed vision for the future.