基于模糊规则的多分类器融合

用非线性方法解决多分类器融合问题能够取得比较高的识别率,但是,当前被应用在多分类器融合领域中的非线性方法可理解性较差,给使用者带来一定的困难.而基于模糊规则的模式识别方法是一类可理解性好的非线性方法,但迄今为止还没有被应用于多分类器融合问题中.基于上述考虑,该文将模糊系统应用到多分类器融合中,并且研究了如何设计可理解性好、精度高的模糊系统的问题,提出了一种改进的基于支持向量的模糊系统设计方法.该方法在从ELENA项目数据库和UCI数据库中选出的4个数据集上进行了测试.实验结果表明,该方法能够用可理解性好的模糊系统实现低错误率的多分类器融合.

Fuzzy Rule-Based Multiple Classifier Fusion

Nonlinear methods perform well in the multiple classifier fusion. However, the nonlinear methods used for the multiple classifier fusion have poor comprehensibility. As a nonlinear method, the fuzzy rule-based pattern recognition has good comprehensibility, but has not been applied to the multiple classifier fusion. Therefore, this paper introduces fuzzy system to the classifier fusion, where the designing issues for accurate and comprehensible fuzzy system are studied, and an improved support vector based fuzzy rule system designing method is proposed. Experiments have been carried out on four data sets from the ELENA project database and the UCI database. The experimental results show that the proposed method can fuse multiple classifiers with low classification error rate based on comprehensible fuzzy systems.