Research On Urban Function Area Recognition Based On Mobile User Behavior Analysis

Research on urban functional area recognition based on mobile user behavior analysis is a key element of mobile user behavior analysis.Comparing to the urban functional area recognition research,this study based on users' behavior data and regular pattern to identify user groups.The next step is to recognize urban functional area by user group categories and distribution pattern.One of the difficult points of this research is how to classify users according to the users' behavior data.The other one is how to classify base stations according to the users' distribution rules and user category label to recognize urban functional areas.There are two researches have been done based on these points in this thesis:Firstly,a dynamic clustering algorithm based on user behavior interest similarity has been imporved in this thesis.The mainly problem how to dynamically divide users according to the user's behavioral interest could be solved by improving the traditional Fuzzy C-Means(FCM)clustering algorithm.The implementation steps are as follow: firstly,the initial clustering parameters of fuzzy clustering were obtained by using Self-organization Feature Map(SOM)Neural Network.Then defining the users behavior interest indicators to measure the user interest in a business behavior,and using the user behavior interest similarity algorithm to calculate the behavior of interest similarity between users as similarity metrics of FCM algorithm;finally,by setting the user average membership threshold to judge whether the clustering result is reasonable.The experiment results show that this method could be better than the traditional fuzzy clustering algorithm to classify mobile users.Secondly,urban functional area recognition algorithm by user behavior analysis based on above research has been imporved in this thesis.This algorithm mainly solves the identification of urban functional area by base station user attribute and it can be divided into four parts: the first step is using the idea of Voronoi polygon to divide map by base stations and according to the time window sequence statistics user's characteristic distribution of base station.Then using the Principal Component Analysis(PCA)method to preprocess the feature matrix;constructing and training the Convolution Neural Network(CNN)classifier based on time window sequence filter,and classify the base station using this classifier is the third step.Finally,by classifying the base station attributes,the base station coverage area is identified to achieve the purpose of identification of urban functional areas.The experiment results show that this method could be better way than traditional urban functional area recognition algorithm to classify mobile users.