Designing And Implementing A Social Search System Based On Clustering Friends In LBSN

The development and popularization of online social networks(OSNs)have brought great convenience to people's daily life.Nowadays,there are billions of users being active in OSNs every day and producing a great amount of social information.Gradually,people transfer their search habits from traditional searching on the webpage like Google and Baidu to searching on the OSNs,social search comes out under such conditions.Traditional search has many drawbacks like low accuracy,longtime to find out,the same results and so on.Under the condition of personalized search,based on traditional search principles,combining social information produced by users,social search could produce personalized search results to improve the quality.In other words,social search could make users find out right people(friends,other similar users or domain experts)quickly and accurately to answer questions.With the appearance of LBSN(location-based social network),after getting the support of mobile devices like mobile phone,tablet and mobile technologies like GPS,WiFi,which could make the foundation to research on the mobility of social search.However,current researches on social search based on LBSN have many drawbacks.For example,social features designed by researchers are not representative and the search algorithms are not efficient.This paper designs multiple social features based on location and relationship information extracted from a LBSN-Foursquare,puts forward a KNN search algorithm based on clustering users' friends and implements a search engine based on inverted index,combining distance to produce high-level results and enhance search speed.To get more accurate search results,firstly,I cluster users' friends.Because LBSNs belong to heterogeneous networks,the dataset is quite sparse.I could make the data more dense by clustering,which aims to reduce the bad influence on search results.Secondly,to design a better algorithm,in addition to traditional social influence,I take professional relevance and distance into account.Overall.I consider search score.social score and distance score comprehensively.In this paper,the dataset is extracted from a real LBSN-Foursquare.After processing.I pick out the data in the New York.I compare another three algorithms with my algorithm,which aims to display the algorithm could improve search results.Finally.I implement the prototype system of social search based on B/S framework.