Understanding Network User Behaviors Based On The Bus Wi-Fi Data

With the continuous advancement of smart city construction,more and more enterprises combined the increasingly mature Wi-Fi technology with public transport,and strive to expand the applications and services for bus travel.At present,there is no research on the network user behavior in the Wi-Fi scene.By studying the access data of the bus Wi-Fi network users,we can obtain the main behavior patterns of the users under the bus Wi-Fi scene.It can help to understand and describe the bus Wi-Fi network user behavior characteristics,applied to the user's differentiation services and advertising precision pushing,also to optimize the planning of bus Wi-Fi network and improve the stability of the network,etc..The research of this thesis relies on the data analysis and mining module in the project of building bus Wi-Fi large data platform,which is presented by the Shenzhen Institutes of Advanced Technology,Chinese Academy of Sciences.It is aimed at mining the main behavior patterns of users by analyzing the data of online time and access content of bus Wi-Fi network based on the built Hadoop platform.The specific work of the thesis is as follows:(1)The behavior patterns mining of users' online time has been actualized.By using the similarity measure method,we revealed and verified the close-similar property of bus Wi-Fi network user online time distribution based on the analysis of the actual user's time usage data.Based on this rule,the hierarchical clustering algorithm is used to mining the user's online time behavior patterns.In order to overcome the shortcoming of too long calculating time in the case of large amount of data,the hierarchical clustering algorithm has been improved in two aspects,one is preliminary grouping of users based on the rule of close-similar,and the second is to merge multiple similar single point clusters at once.Finally,four kinds of behavior patterns from the data are obtained.According to the actual application scene,the improved algorithm is evaluated from four indicators.The results show that the improved algorithm has obvious advantages in distinguishing the time characteristics of different patterns and the operating efficiency.(2)The behavior patterns mining of the user's surfing content has been actualized.In view of the difficulty of effectively processing and expressing the user's behavior information for the complex and diverse bus Wi-Fi user's surfing data,and based on the analysis of the Internet access characteristics by intercepting the secondary domain name of the user's access URL,this thesis presents a solution: according to the number of access users to filter the sites with search significance,then,classifying the site using the app classification method with the mobile Internet report.And the rationality of this solution is verified by the real scene data,construct the feature vector of user's online content analysis.And then,based on the characteristics of data sparseness and similar user interests,a behavior patterns mining model of users' online content based on weighted is proposed.The model uses TF-IDF to weight the feature,SVD to dimension the sparse matrix,and then uses the improved K-means algorithm to extract the nine kinds of behavior patterns of users' surfing content.This thesis evaluates the model from four indicators.The results show that the model proposed in this thesis is effective in mining the behavior patterns under the bus Wi-Fi scenario.(3)The user behavior analysis system for bus Wi-Fi network has been designed and achieved.The framework of the system is designed,and the design and implementation of data access,data preprocessing,data mining and data storage are introduced in detail.And the results of the system are demonstrated by visualization,the application of the system is also presented in the thesis.