异构信息网络的分类研究

ON CLASSIFICATION OF HETEROGENEOUS INFORMATION NETWORKS

异构信息网络在生活中无处不在,并且逐渐起着非常重要的作用。如何在信息网络里分类这些节点,对节点进行排名,已成为研究的热点问题。结合排名和分类方法,提出一种新的基于排名的迭代分类框架,建立一个基于图形的排名模型,使计算的排名分布在每一类的对象中。在每次迭代中,根据当前的排序结果,排名算法中使用图形结构的调整,使子网对应于特定的类被强调,而网络的其余部分被削弱。实验证明,整合分类排名不仅产生更准确的数据分类方法,而且在每一类中,能提供有意义的对象排名,比传统分类更有效。

Heterogeneous information networks exist everywhere in real life, and gradually play very important roles. How to classify these nodes in information networks and to rank them has become a research focus. By combining ranking and classification methods together, in this paper we present a new ranking-based iterative classification framework, build a graph-based ranking model, and make the calculated ranks distribute in every type of object. In every iteration, according to current ranking results, in ranking algorithm the graph structure adjustment is Used, this makes the subnets corresponding to a specific class are emphasised, and the rest of the network is weakened. Experiment proves that the integrated classification ranking not only results in more accurate data classification method, furthermore, in each category it can provide meaningful object ranking as well, which is more effective than traditional classifications.