数据分割的分子动力学算法

数据分割研究的基本内容是数据的分类和聚类,是数据挖掘的核心问题之一,在实际问题中应用广泛.特别是针对有向网络数据的研究更是学科发展的前沿.但由于这类问题结构的非对称性,使得模型与算法的构建存在本质困难,因此相应的研究结果较少.本文借鉴分子动力学方法的思想,提出了一类新的网络数据半监督分类模型及算法.该算法不仅适用于关系对称的无向网络数据,而且适用于关系非对称的有向网络.最后针对期刊引用网络数据进行了数值实验,结果表明了模型及算法的可行性和有效性.

MOLECULAR DYNAMIC ALGORITHMS FOR DATA PARTITION

Data partition models such as clustering and classification are crucial in data mining and are widely used in many fields of application. Specially, data partition on directed network is an important issue. The nonsymmetric structure brings the essential difficulties to the modeling as well as the algorithms. Based on the molecular dynamic methods, we proposed semi-supervised classification models and algorithms for network data, which could be applied to not only undirected network but also directed network. Journal citation web is used as the data set to test the models and algorithms in this paper. The numerical results show the feasibility and effectiveness.