Theoretical Research And Application Of Big Data

In recent years,research and application of big data(Big Data)has been aroused wide attention of academia and enterprises,Governments have also given sufficient attention and support from the strategic and policy levels,Coupled with the rapid development of Internet technology,and jointly promote the theoretical innovation,technological innovation of big data.As is known to all,human society has entered the DT era from the IT era,and big data has played an important role in the fields of economy,military,education,environment,agriculture and transportation etc.especially,the financial,IT and communications industries,which are closely related to our lives,are models of big data applications.This paper summarizes the research backgrounds and the development status of big data at home and abroad.The basic concepts of scientific big data are described in detail,including data source and data feature,big data four-step processing flows.And this paper starts from the modern communication enterprise,by using the theoretical basis and technical system of Big data to make an empirical study on the behaviors of on-line users and off-line users in the communication industry.In particular,using the big data processing methods to analyze the data of post-paid mobile users to build mobile users' on-line and off-line identification model.forecasting users' off-line risks by using real-time data stream of mobile users.The main contributions of this dissertation are follows:(1)In the context of this study,after several scientific experiments,the optimal data mining algorithm—RBF(Radical Basis Function)artificial neural network is selected to construct the mobile user's off-line behavior classification and recognition model.(2)Based on a method of input output correlation(IOC),In this paper,we give the class dichotomy method to choose the optimal combination of attributes.This method not only solves the problem that the optimal combination of variable attributes can not be selected in the process of RBF artificial neural network modeling optimization,but also accelerates the speed of selecting the optimal variables compared to the methods of Wen Zhuan[7] et al,thus improving the efficiency of the model.Finally,the research work of the full text is summarized,and the key technology in the process of post-paid mobile user behavior data processing is prospected.