基于位置的社交主题推荐模型

Social Theme Recommendation Model Based on Location

针对社交网络以及社交用户关注主题,分析用户所在位置,在社交网络的基础上提出主题推荐模型即location-themesocial model(LTS M odel).文章主要从三个方面进行了分析,首先对主题进行分类,运用余弦相似性算法构建向量空间主题模型.其次,在MapReduce框架下根据位置快速构建R*-tree索引,建立空间模型,在此基础上找到基于位置和主题的社交网络模型.最后,使用标准数据集对算法进行测试,并根据准确率、召回率和F1值对其效果进行评价.R*-tree索引算法采用抽样方法快速确定空间划分函数,保证了数据对象均匀地划分到各个分区.余弦相似性算法能够快速准确地找到相似主题,并且敏感识别度较强.实验证明基于LTS Model的位置—主题推荐算法(LTRA)能够快速找到满足用户兴趣的主题并进行推荐.

Location-theme-social model is put forward by considering the social network and analyzing where the user is and what themes they focus on. The thesis mainly analyzes it from three parts,firstly ,categorizes the topic of an input and constructs vector spa- tial theme model by using cosine similarity. Secondly, creates R ＊ -tree index based on location under the framework of MapReduce and constructs spatial model, so that to find the social network based on location and theme. Lastly, tests the effects of the algorithm with standard data sets which is evaluated by precision, recall and F1-Score. R ＊ -tree index constructed with the method of sampling could quickly determine the space partition function and ensure that the object evenly divided the space to the various partitions. Cosine similarity could find similar themes quickly and accurately. It has a strong sensitive recognition. Experiment has indicated that the locationtheme recommendation algorithm based on LTS Model could find the interested themes for users and recommend them.