基于FRS-SVM的工业数据挖掘研究

Research of the industrial data mining based on FRS-SVM

在工业生产分类过程中,存在数据量过大,冗余特征过多的问题,本文结合模糊粗糙集和支持向量机研究了一种分类算法。首先采用模糊粗糙集方法对条件属性进行属性约简,找出对分类决策具有主要影响的特征。以约简结果作为分类模型的输入变量,然后利用支持向量机对样本进行训练,建立分类模型,最后将本文的方法用于地板正反面分类和分析氧化铝晶种分解过程,并测试模型的分类效果。MATLAB仿真实验的结果表明本文的方法是有效的,具有分类正确率高,结构简单,泛化能力好的优点。

A classification algorithm is studied based on the fuzzy rough set and support vector machine about industrial processes which has too much redundant features and large amount of data in this paper. Condition features are reduced by fuzzy rough set, and fred the characters of has a major effect on classification of the floor. With the reduction result as input variables, SVM is used to train samples, establish classification model, then this method is applied to classification of the floor and alumina seed decomposition process, also test classification results. MATLAB Simulation experiment results show that the method is effective, high classification accuracy and has the advantages of simple structure, good generalization ability.