面向社交媒体嵌入关系数据感知方法的研究
A Research on the Data Aware Method for Social Media with Embedding Relationship

针对社交媒体数据感知成本高、数据感知效率低等问题,提出了社交媒体嵌入关系多阶段数据感知方法(online social media？？multi stage data aware, OSM？？MSDA)。该方法以数据感知对象内部关系的分布特征为基础,构造一个具有偏好特征的种子网络;采用Metropolis？？Hastings方法优先选取数据感知对象中高度节点的邻接关系,快速填充特征网络,实现网络轮廓探测;使用基于马尔可夫生灭机制的延迟拒绝方法控制概率转移核,对局部耦合关系进行修剪,确保连通关系疏密的合理分布。实验结果表明:OSM？？MSDA建立的多阶段渐进数据抽样方法,能够克服已有数据感知方法采集样本的盲目性,在宏观尺度准确、高效的感知社交媒体嵌入关系的社会资本特征,确保特征网络与数据感知对象的结构更具有一致性,同时还能降低数据的使用成本,将数据处理效率提高32％~63％。
A multi stage data？？aware method for online social media with embedding relationship (online social media？？multi stage data aware, OSM？？MSDA) is proposed to solve problems of data aware in online social media, such as poor availability, high business cost, and low？？efficiency, et al. A seed network with preference characteristics is constructed, and then the Metropolis？？Hasting method is used to choose adjacency relation with high degree in data aware population. Finally, the improved Delay？？Rejection method is used to regulate the Markov probability transition kernel, and to control the distribution density in local network. Experimental results show that OSM？？MSDA gets more precise results for social capital of social media and high？？efficiency at macro？？level, and overcomes the blindness of existing data aware methods. At the same time, OSM？？MSDA ensures the consistency between the characteristics of network and the structure of the data object perception, reduces the cost to use data, and increases the data processing efficiency by 32%？？63%