Characteristics Analysis Of Mobile Social Network

With the rapid development of Internet technology,the number of domestic network users has reached an unprecedented height.Internet users could access into Internet via fixed or mobile network provided by domestic operators,thus produced various types of traffic data which shows the behavior of Internet users.It has become a very popular topic that how the behaviors of Internet users can be analyzed and produce valuable information efficiently.In this paper,we mining not only the data of user behavior collected by technology of DPI from a large domestic Telecom operators,but also the traffic data in Advertising platform of DSP which can infer the characteristics of users.The conclusions can be acquired from these macro statistics includes behaviors of user groups,interests' item set of users,statistical analysis of large-scale Internet data,and traffic flow characteristics of Internet users.This paper introduces the complete process of the mining of these Internet data.Firstly,we use the Web crawler framework called Scrapy realized by Python to obtain the category labels of Numerous domains and urls,then upload these labels to HDFS in an appropriate manner;Secondly,we set up the Hadoop distributed data processing platform on the Linux,and install several data storage and algorithmic analysis modules including Hive,HBase,Mahout on Hadoop;Thirdly,based on the MapReduce framework,we develop several innovative algorithm including the data preprocessing and label classification module using Java,the algorithm of Secondary Sort,the improved Canopy K-Means clustering algorithm,the improved FpGrowth frequent item sets discovery algorithm,and the self-similarity of traffic discovery algorithm.Finally,we elaborate the analysis of the output data of various provinces and periods in detail in this paper.Considering the part of data analysis in this paper,the overall basic statistical results of these two data sources will be showed respectively.Then,using the form of statistical consultation report,the detailed statistical analysis of domestic E-commerce platform and DSP platform will be presented.Moreover,the data analysis in this paper will involves the explanation of user action sequences,the analysis of the user interests clustering results,the frequent items analysis and the research of self-similar users' traffic.As for the significance of our analysis,the overall basic statistical results will contribute to the environmental observations and regulations of Internet.Then,the digging of users' individual interests in relevance,groups'properties tendency,as well as the traffic self-similarities could benefits in fields of directional recommendation of advertisement,the advancement of browsing experience and the optimization of marketing tactics by several companies.