基于泛组合运算的分类器融合  

Fusion Classifiers Based on Universal Combination Operation

作　　者：林卫[1] 何华灿[1] 贾澎涛[1] 

机构地区：[1]西北工业大学计算机学院,西安710072

出　　处：《计算机科学》2007年第10期204-207,共4页Computer Science

基　　金：本文受国家自然科学基金项目(No.50474041)资助。

摘　　要：本文介绍了一类可用于分类器融合的泛组合逻辑算子,同时作为方法论基础研究了基于案例学习的泛逻辑运算符(算子)构造和选择方法,并以此构造分类器的融合器,实验数据集选择了UCI的spam数据集,并同其他融合方法进行了对比。结果表明本文所述方法具有较低的错误率和相应较高的查全率。In this paper we introduce that combination operator can be a kind of universal logic operator to describe complex system in noisy and changing environment, e.g. the problem of fusing a collection of classifiers. This approach is mainly based on Triangular norm theory. We also discuss the learning method for fuzzy logical operator construction and selection;finally we apply this in fusion of classifiers. The results of computer experiments on spare and other UCI datasets show that the method can give less error rate and better performance correspondingly.

关 键 词：泛组合运算 模式分析 分类器融合 T-范式 泛逻辑学 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]