Analysis Of Network User Behavior Using Wavelet Clustering

With the rapid development of the Internet, the number of network users have reached inconceivably number, variety of network applications become increasingly rich. Internet has become more and more popular in work, entertainment and services.In recent years,network operators and service providers are interested in characteristics of network user behavior,habits and applications of using, and they want to get the user behavior information through some technical means, these achievements can be an important basis for network user management, quality of service optimization and online marketing strategy, even more providing customer relations value-added services for third parties.Network user traffic data is analyzed based on wavelet analysis theory in this thesis,we can get many different behavior patterns of users by clustering. Providing useful data to support user service or network marketing. There are main works including the following aspects in this thesis:①Firstly,abstracting network user behavior and building a network user behavior model. Business behaviors of network users are used as analyzing targets in thesis, in order to simplify the analysis dimension, filtering properties directly related to business practices in many kinds of network traffic. In thesis, model is built through network application,time ratio and consumption flow ratio, this three aspects reflect user behavior characteristics.②Wavelet analysis theory is studied in this thesis, and improving the existence of insufficient of traditional wavelet clustering algorithm high complexity space. An improved wavelet clustering algorithm named Mo FSU is proposed in this thesis. In order to optimizing space complexity and improving time performance, this algorithm reduces the space consumption by merging the feature space. The proposed algorithm is effective by theoretical analyzing in this thesis.③ Theoretical perspectives are verified by experiment in this thesis. Clustering results are obtained by network user behavior model, and improved algorithm through analyzing network user behavior log data set. some other similar algorithms are compared on this data set. The result shows that time complexity and clustering quality have been improved in improved algorithm.