智能参数学习的模糊决策树算法  

Fuzzy decision tree induction based on optimization of parameters

作　　者：孙娟[1] 

机构地区：[1]河北大学数学与计算机学院河北省机器学习与计算智能重点实验室,河北保定071002

出　　处：《计算机工程与应用》2012年第23期148-154,共7页Computer Engineering and Applications

基　　金：国家自然科学基金(No.10804025);河北省自然科学基金(No.F2010000318)

摘　　要：模糊决策树算法在处理数量型属性的数据时,需要进行数据模糊化预处理。但是,每个数量型属性应该模糊化为几个语言项通常要凭经验设定的,目前还没有使用标准粒子群优化算法(PSO)自动设定语言项个数的研究。提出使用PSO确定语言项个数的模糊决策树算法(FDT-K算法),通过实验证明FDT-K算法产生的模糊决策树性能明显优于凭经验设定语言项个数所产生的模糊决策树。Fuzzy Decision Tree induction（FDT）has been used in more and more application area.When the data are numerical,the FDT algorithms need to fuzzify them into some linguistic items.That how many linguistic items of an attribute are proper is not known.The selection generally depends on experts’opinion or people’s common.Currently,it is not yet available to learn the number of linguistic items by using the experimental method of particle swarm optimization.The paper introduces a PSO based approach to optimize the selection of linguistic item’s number in fuzzifying processing of data in FDT（FDT-K algorithm）.Experimental studies show that the FDT-K algorithm compared with the people’s common methods to decide the number of linguistic items of attribute can create a better fuzzy decision tree with higher classification and generalization capability.

关 键 词：归纳学习 模糊决策树 数据预处理 模糊化 粒子群优化算法 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]