User Behavior Analysis Based On Operation Data Of Network Service Provider

Currently,operators are facing a more complex and fierce competition.First of all,with the rapid development of the mobile Internet,traditional communication operators are gradually at a disadvantage of being pipelined and marginalized.Secondly,the decline in voice traffic and the high costs of construction and maintenance brought by the rapid growth of traffic have also slowed the growth of operators'profits.To this end,user-centric data analysis and mining can help operators cope with the above-mentioned competitive situation.Among them,identifying the important location of the user(the location of frequent visits)can not only provide users with personalized network services,but also help urban planners optimize various resource allocations of the city and reduce network investment costs.Analysis of association rules can help operators to conduct refined marketing for users with different consumption characteristics,improving users' satisfaction and the overall revenue of operators.This paper mainly uses the user log data of the operator to study the user's behavior.Firstly,it analyzes different types of user data.Then it identifies the user's residential location and working position.Finally,it analyzes the association rules based on the user's consumption.The main contents of this paper are as follows:(1)Do statistical analysis based on operator log data.Statistical analysis is performed on user call log record data,user online log record data,cell and grid association data,user attributes and consumption data.Understand the overall distribution of data,and make preparations for user's important location identification and association analysis based on users'consumption characteristics.(2)Propose a method for user important location identification using a space-time dimension.The existing method based on cluster analysis for important location identification only considers the time dimension.The proposed method adds the spatial dimension of the area marker to which the base station belongs,and uses the spatial and temporal dimensions to perform important location identification.Finally,the effectiveness of the proposed method is verified indirectly.The results show that compared with the original method,the method further improves the accuracy of the user's important location identification.(3)Propose a user consumption feature correlation analysis based on user communication log data.Firstly,the user log data is mined from the time dimension and the spatial dimension,and the feature extraction is performed,which includes:using the clustering method to discover the behavior mode of the Internet/call of the user at different times,calculating the cumulative moving distance,the average circle radius,the number of places visited and commute distance.Then,we use k-means to mine user consumption record data and extract consumption characteristics.Finally,we use FP-growth algorithm to mine the potential relationship between user consumption level,space-time behavior characteristics and user attributes.And get seven effective strong rules to provide reference marketing suggestions for operators on the basis of the rules.