Multi-source Based Personalized Location Recommendation Services Research

As environmental sensing and computing technologies developing a wide variety of big data began to appear in the environment of urban computing.The many statics and problem of the city can be analyzed and get the answer from these big data.The many characteristics and problem of the city can be analyzed and get the answer from these big data,this is the concept of the urban computing.By Integration,analysis and calculation the alot of data generated by the moving terminals,we can find many knowledge and laws.And bu using it,to improve the efficiency of the co-working of human environment and city.Users in LBSN is moving continuously,there position is still changing.An effective index is needed to acquiring and manging these moving objects effectively.At the meaning time,mmoving objects in the city environment usually carring more information rather then only spatial-temporal information,so this paper proposed a new kind of spatial-temporal index ATPR-tree.This kind of index carring the attribute dimension.Can managing the objects with additional attribute value.TPR-tree's structure is the basic structure our ATPR-tree.But because of the new dimension has been added.so we need to amended the TPR-tree's structure.The new added dimension definitely changed objective function.And our work also amended the insertion,reinsert and the deletion algorithm of the orignal TPR-tree by this new objective function.The experiment part will compare the efficiency of the TPR-tree which didn't added the attribute dimension and our ATPR-tree.The experimental data is generated by the GSTD.The experiment result shows ATPR-tree can improve the querying effciency effectively.The origanl location baesd recommendation system using the similarity users and the local experts' preference to make location recommendation.But in this process,the data we used is always obtained by the user' trajectory data that has been already generated,and ignore the factors that are changing in real time.So there will be some error in the recommendation results.To solve this problem,we propose the concept of the modified set.This modified set will refer to the current state of the LBSN users.Recommendations is made not only based on the users' s trajectory that has been already generated and also based on the feedback of users' current positions,and by this way to make more precise recommendations.At the last,we testing the effectiveness of modified set to make the recommendation more precise,by a series of experiments.