基于模糊积分和生态捕食模型的多分类器融合  

Multiple classifiers combination based on fuzzy integral and prey-predator model

作　　者：黄晶[1] 姚跃华[1] 周杰[1] 

机构地区：[1]长沙理工大学计算机与通信工程学院,湖南长沙410077

出　　处：《计算机工程与设计》2008年第23期6066-6068,共3页Computer Engineering and Design

摘　　要：考虑到各个独立分类器之间的重要程度和相互作用,根据模糊测度Sugeno积分的理想特性,用模糊测度代替各个分类器的权值,利用捕食-被捕食模型动态周期性变化的特点,将模糊测度对应于种群规模并不断随之进行调整。通过仿真实例证实了新的多分类器融合模型具有较低的分类错误率,能够有效地提高分类精度。Considering the difference of mutual affects and important degree between each independent classifier, a new multiple classifier model is presented. Fuzzy measure is used to instead of the weights of classifiers according to ideal performances of fuzzy measure refer to Sugeno fuzzy integral. Because of the periodically dynamic attributes of prey-predator model, fuzzy measure corresponding to the scale of population is adjusted. The experimental results show that this new model efficiently improve the classifier accuracy.

关 键 词：模糊测度 捕食-被捕食 周期性 种群规模 多分类器融合 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]