基于通讯数据的社群分类

发布时间：2018-06-28 10:22

本文选题：马氏链 + PageRank算法　；参考：《华中师范大学》2017年硕士论文

【摘要】：近年来,对研究对象的分类问题已经在多个研究领域得到了广泛的应用,分类的方法也得到了巨大的发展,如聚类分析、KNN算法、决策树、支持向量机等等。本文以研究实际通讯数据的社群分类问题出发,提出了一种PageRank算法和SimRank算法相结合的新的分类方法。在两个实际案例中得到实践,并将结果分别与真实情况和传统聚类方法结果进行了比较,整体效果和结果解释均较为理想。本文提出的方法适用于研究任意对象与对象之间的关系。先将问题转化为一个简单而直观的点边结构图模型,通过PageRank算法计算状态点在整个图中的“重要性”,通过SimRank算法测量对象之间结构上的相似性,根据它们与其他对象的关系,有效地进行分类。本文的基本思想是“两个对象是相似的,则与他们相关的对象应相似。”本文的研究可以视为无监督学习(无指导学习)的实践和探索。
[Abstract]:In recent years, the classification of research objects has been widely used in many research fields, and the classification methods have been greatly developed, such as clustering analysis KNN algorithm, decision tree, support vector machine and so on. In this paper, we propose a new classification method which combines PageRank algorithm with SimRank algorithm. The results are compared with the real cases and the traditional clustering methods respectively. The overall effect and the interpretation of the results are satisfactory. The method proposed in this paper is suitable for studying the relationship between arbitrary objects and objects. Firstly, the problem is transformed into a simple and intuitionistic point-edge structure graph model. The importance of state points in the whole graph is calculated by PageRank algorithm, and the structural similarity between objects is measured by SimRank algorithm, according to their relationship with other objects. Classify effectively. The basic idea of this paper is that "if two objects are similar, the objects related to them should be similar." The research in this paper can be regarded as the practice and exploration of unsupervised learning (unsupervised learning).
【学位授予单位】：华中师范大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：C815

【参考文献】