Using Decision Boundary to Analyze Classifiers

摘要：

In this paper we propose to use decision boundaryto analyze classifiers.Two algorithms called decisionboundary point set(DBPS)and decision boundaryneuron set(DBNS)are proposed to obtain the data onthe decision boundary.Based on DBNS,avisualization algorithm called SOM based decisionboundary visualization(SOMDBV)is proposed tovisualize the high-dimensional classifiers.The decisionboundary can give an insight into classifiers,whichcannot be supplied by accuracy.And it can be appliedto select proper classifier,to analyze the tradeoffbetween accuracy and comprehensibility,to discoverythe chance of over-fitting,to calculate the similarity ofmodels generated by different classifiers.Experimentalresults demonstrate the usefulness of the method.

作者: Zhiyong Yan Congfu Xu

作者单位: College of Computer Science,Zhejiang University,Hangzhou,China

会议类型: 国际会议

会议名称: 2008 3rd International Conference on Intelligent System and Knowledge Engineering(第三届智能系统与知识工程国际会议)(ISKE 2008)

会议地点: 厦门

会议语种:英文

页码: 302-307

在线出版日期: 2008-11-17（万方平台首次上网日期，不代表论文的发表时间）