Abstract: An evolutionary approach for designing compact fuzzy classifier directly from data without any a priori knowledge of the data distribution is proposed. The variable input spread inference training (VISIT) algorithm is used to create each individual fuzzy system, and then searches the best one via genetic algorithm. Rules and membership functions are automatically created and optimized in an evolutionary process. In order to effectively evaluate the accuracy and compactness simultaneously, a fuzzy expert system acts as the fitness function. The experiments on two benchmark classification problems show the effectiveness of the new method.

Key words: VISIT, Genetic algorithm, Compact fuzzy classifier, Rule extraction

摘要： 介绍了在没有数据分布先验知识的情况下，用进化方法直接从训练数据中建立紧致模糊分类系统的方法。使用VISIT算法获取每个个体模糊系统，再用遗传算法从中搜索最优的模糊系统。规则和隶属函数是在进化过程中自动建立和优化的。为了同时有效地评价系统的精度和紧致性，用一个模糊专家系统作适应度函数。在2个基准分类问题上的实验结果表明了新方法的有效性。

关键词: VISIT, 遗传算法, 紧致模糊分类器, 规则抽取