Research And Application Of Clustering Algorithm Of Big Data In Mobile Social Networks

With the rapid development of mobile social networks,because the ubiquitous GPS device provides a data base,personalized user location services will also become a hot research.Essentially,the nature of the user’s movement is to move the location of a timestamp with the <location,time>,this thesis propose the GSRM(Geographic Service Recommender Model)and Geographical Location Prediction algorithm based on Joint of Distance decreasing function and prefix-Tree structure(short for GLP-JDT).The three main works of this thesis are as follows:(1)A whole frame for GSRM.GSRM is a complete framework for clustering based location services.First,from the original data set,we use the “Stay Point” algorithm to extract the user’s stay point,and filter out most of the GPS points without semantics.Because this thesis without check-in data for location services,so we use "folk wisdom" method.For this reason,the system clustered all the users’ stay points to find hot points which are public are recognized.Then,based on the consideration of large data,the clustering algorithm is transplanted on the distributed platform with the help of the MapReduce framework of Hadoop to improve algorithm efficiency.Futermore,MiningMP algorithm is used to mine the user’s trajectory sequence to find the frequent sequence patterns.Finally,we used GLP-JDP algorithm to predict the location of the user,we use the Baidu map API to recommend the services to the user.(2)GLP-JDT algorithm based on Clustering.As an application of clustering analysis,we can predict the location of the next location of the user,and we propose the GLP-JDT algorithm based on clustering.For the new users who do not have tarce mode in postion preditor,our GLP-JDT algorithm sets up a cold start to predictor.Then,based the feedback information of users,scoring system recalculates the position prediction parameters,and constantly improves the range of parameters through repeated iteration.Finally GLP-JDT predicts a current optimal location.(3)Evaluation of GSRM system and GLP-JDT algorithm.In order to verify the GSRM model and the GLP-JDT algorithm based on clustering,experiments were conducted based on real GPS trajectories from Microsoft Research Asia(20 million real trajectory data).The experimental results clearly demonstrate that our proposed GSRM model is effective and efficient at predicting locations.In the next point prediction,the GLP-JDT algorithm can be applied to a variety of user groups;and in terms of time and space effects,the average accuracy is higher than the benchmark model.