With the rapid development of Internet and smart phones,people are more and more inclined to use the mobile phone to access the Internet and realize various needs through APPs.By mining the associations behind APPs,we can discover potential APP usage patterns that help with relevant system applications and network optimization.Based on the mobile operators' DPI data,this thesis extracts the large-scale data of users using APPs,mining association rules,and finally gets the association behind APPs to support user behavior mining.Specifically,the main work of this paper includes:1.We distributed processing platform to complete the preprocessing and storage of user data,including APP traffic features collection,DPI data matching and filtering,data merging based users,etc.On this basis,we design a framework for the association analysis of APP,which includes data collection,data processing,data analysis,association rule mining algorithm and rule extraction.2.This thesis proposed an improved algorithm of association rule mining,DPI-MSFUP.It not only deals with the problem of rare items,but also realizes incremental updating.The effectiveness of DPI-MSFUP algorithm is verified by experiments,and the time complexity is improved.Compared with the typical Apriori and MS-Apriori algorithms,this algorithm is more suitable for large-scale APP association rules analysis.3.We use the DPI-MSFUP and other algorithms to dig out the association rules of mobile APPs and found that the mobile APP association rules can be divided into four categories:similar competition,similar complementary,heterogeneous competitionand heterogenous complementary.And the rules are time-sensitive,extensive,transformable,diversity and no transitive. |