不确定性研讨信息的结构化与融合方法

基于流形正则化框架提出一种分类算法(MLD-RLSC),以解决高维文档分类问题。该算法通过构建训练样本的最近邻图来估计数据空间的几何结构并将其作为流形正则化项,结合多变量线性回归获得高维文档的低维流形结构,并采用k近邻分类器对低维流形进行分类,得到针对多类问题的分类器。该算法能够充分利用训练样本的类别信息来帮助学习以提取有效特征。通过在Reuters-21578数据集上的实验,证明该算法的分类性能和运行速度比传统分类器有较大的提高。

Method of Structure and Fusion for Uncertainty Seminar Information

In group discussion, because the thinking of experts is uncertainty and information is unstructured, group discussion is difficult to reach consensus. To solve this problem, this article gives a information model which is compositedof natural and artificial property. Mechanism of how to extract uncertainty information is studied with argument frame-work,valid arguments group and distribution function. A method is discussed,which is information fusion based on a-verage argument. All is to promote awareness of the spiral and groups to reach consensus. Finally, a case is provided toprove feasibility and effectiveness of model and method,which is discussion about customer's demand of car.