基于泛组合运算的分类器融合

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。

Fusion Classifiers Based on Universal Combination Operation

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 ap- proach is mainly based on Triangular norm theory.We also discuss the learning method for fuzzy logical operator con- struction and selection;finally we apply this in fusion of classifiers.The results of computer experiments on spam and other UCI datasets show that the method can give less error rate and better performance correspondingly.