Optimization Of Mobile Internet Traffic Collection System And Analysis Of User Mobility

Today, with the rapid development of mobile Internet, the mobile user behavior analysis and QoS(Quality of Service) improvement have become a hot topic of mobile Internet research. According to the characteristics of mobile Internet data, analysis of user mobility will be of great help to improve the quality of service of mobile Internet and analyzing user behavior specifically. We need to collect high-speed packets of the mobile Internet, from which to extract the required fields and to carry out the relevant data analysis and application development.Firstly, this thesis analyzes the basic framework and data structure of mobile Internet traffic collection system, and briefly introduces the relevant technology of mobile Internet traffic collection. Through experiments, the performance of several queue models and memory pool models are analyzed in detail. After that, the background of the traffic distribution system is introduced, and several common distribution algorithms are listed. Through experiments, the performance of different distribution mechanisms and algorithms are analyzed.In the mobility analysis section, we firstly introduce the Hadoop software framework which can deal with the massive data efficiently and rapidly. As a large data processing framework, Hadoop can be applied to process and extract information of user behavior from the traffic data of mobile Internet. Secondly, the application of mobility research is briefly introduced. At the last part, because the analysis of user mobility behavior has a lot of valuable applications, we study some location prediction algorithms and the user relationship discovery algorithms. We introduce three algorithms that can be used to predict the location of mobile users:LZ-Based algorithm Markov algorithm and Improved Markov algorithm.After the introduction to the principle of the three algorithms and extracting the path of mobile users, computational complexity and accuracy of the two algorithms are compared on the basis of forecast data obtained through experiments. We find that the Improved Markov algorithm has a better performance than the other two.