Design And Implementation Of Friend Recommendation System Based-on Interest

With the popularization of the Internet,the social communication mode between people and people has gradually moving from the traditional social communication mode to online social communication mode.For this reason,people try their best to develop more informative and more powerful social networking tools,in order to satisfying our growing demand of online social communication.Our society is an open society,which needing the open social communication,so in the online social communication,users hope to set up their friends group in the social networking tools,try to find some people who having the same interests to share their ideas with them.Therefore,if we want to implement an efficientive and accurate friend recommendation system,we have to dig out the effective information in the users,personal information and their publishing messages.This thesis considers large number of social network data as the research object,designing and implementing a friend recommendation system based on interest.The implementation of this system is based on MapReduce computing framework,using the distributed file system(HDFS)to store the data.The core function of this system includes three parts:The first is the distributed storage management of social network data,the main implementation of this modular includes uploading data,downloading data,deleting data,creating new folder,renaming the data,which is stored in HDFS.This modular can facilitate the administrator to update and maintenance the social network data.The second is the friend recommendation based on interest.Use the K-Means clustering algorithm for cluster analysis of social network data and use the Bayesian method for classification of social network data,in order to implementthe friend recommendation based on clustering and classification.The third is improving the users' interest tags.Using these two methods of data mining to dig out the hidden interests of social network users and improve the personal information of them.This thesis uses large number of micro blog data to test the system,including the module testing and the function testing.The results show that the system is a high-reliability and high-safety system,has a reasonable response time,can implement an efficient and accurate friend recommendation function,and can provide an intuitive and friendly interface for the user.In this thesis,we firstly introduce the status and related content of friend recommending research.Then,in order to clear the function and feasibility of the system,we analyze the system demands from two aspects:demand and feasibility.Then,we carry out the design of the system including the overall design and module design.After that,we carry out a detail design and implementation for the system,and give the specific design and implementation steps.After the implementation for the system,we test the system including the unit test,integration test and the function test.Then we demonstrate and evaluate the test results.Finally,we summarize the characteristics and shortcomings,and pointed out the direction of future work.