Spatio-temporal Analysis And User Interest Mining Based On Cellular Network Data

With the popularization of 3G and 4G networks,the mobile Internet has been rapidly developed.Along with the maturity of the mobile terminal market and the introduction of new smart terminals,the number of users in cellular networks has also been expanding.Data services in cellular networks have become an important research object for people.With the increasing data traffic of cellular networks,how to analyze and study the massive cellular network data using big data technology has become a hot research topic nowadays.On the one hand,it will effectively promote the planning of the existing cellular network construction and improve the performance of the network from the perspective of business;on the other hand,the data mining through the cellular network implicitly and usefully Information,has become an urgent need for business,engineering.In this paper,we set up a big data platform of cellular network to study the business data of cellular network.This paper analyzes the spatio-temporal patterns of urban snapshots under large-scale,and at the same time,excavates the user's search interests on a small scale.The main work includes:1.Based on the existing big data technologies such as Hadoop and Spark,combined with the processing requirements of data services in cellular networks,a set of big data analysis and processing platform for data analysis of cellular networks is built.Then we introduce 4 core modules of the platform including data acquisition module,data storage module,data processing module,data display module.Finally,we describes the key problems encountered in the big data system and the solutions.2.Based on instant messaging traffic,this paper presents a city's business snapshot mode analysis and research methods.Firstly,a method of establishing business snapshot is proposed.Based on the concept of image similarity measurement in image recognition,a snapshot measurement method of S-PSNR and S-SSIM is proposed.Then using the unweighted pair-group method with arithmetic means to cluster the snapshots,it is found that compared with the traditional space-time model,the proposed urban snapshot space-time model can get a more accurate behavior pattern of city working day and weekend.Finally,the pattern comparison analysis of the clustering results among cities with different traffic volume reveals that the urban users with high traffic volume have more obvious daily behavior patterns.3.Based on search business of the cellular network,we describe the user search interest mining and interest forecasting.First of all,this paper proposes a lexical resolution scheme for searching data in cellular networks.Based on the search terms,a special short text classifier is established to map the search terms.Then the search interests are spatiotemporal analyzed,which can be divided into three categories according to the search volume in time.The spatial search interest is closely related to the actual geographical POI.At last,the trajectory of the search interests is established,and an interest pattern mining algorithm based on Apriori is proposed.Based on proposed pattern,an interest prediction algorithm is proposed.The result show that compared with the interest prediction algorithm based on first-order Markov model,our proposed algorithm is improved by 18.57%on average.