一种对象完备度优先填补的决策树规则提取算法  

A DECISION TREE RULES EXTRACTION ALGORITHM WITH IMPUTATION PRIORITY IN OBJECT COMPLETENESS

作　　者：陈家俊[1] 苏守宝[1,2] 金萍[1,3] 

机构地区：[1]皖西学院信息工程学院,安徽六安237012 [2]金陵科技学院,江苏南京211169 [3]中国科学技术大学计算机科学与技术学院,安徽合肥230027

出　　处：《计算机应用与软件》2014年第5期264-267,294,共5页Computer Applications and Software

基　　金：国家自然科学基金项目(61075049;61375605);安徽省高校自然科学研究重点项目(KJ2012A274)

摘　　要：不完备信息系统中决策规则的提取是数据挖掘领域的重要研究问题。对不完备信息系统中决策规则的主要获取方法进行分析,以决策属性具有缺失值的不完备决策表为研究对象,提出一种基于数据优先填补的决策树规则提取算法。针对ROUSTIDA算法在数据填补时运算量较大且容易导致决策规则冲突这一问题,算法采用决策属性优先填补的思想,引入对象完备度概念对其进行改进,使用改进的ROUSTIDA算法对不完备决策表进行一次性数据填补预处理,并在限制容差关系下采用属性重要性为启发函数构建决策树,从而获得决策规则。实例表明该方法是有效的,生成的决策规则简单,且具有较高的精确度。Decision rules extraction in incomplete information systems is an important issue to be studied in data mining field. We analyse the principal decision rules acquisition method in incomplete information system,and take the incomplete decision table with missing decision attribution values as the research object,propose a data imputation prior-based decision tree rules extraction algorithm. For the deficiency of ROUSTIDA algorithm that it has large amount of computation in data imputation and is easy to cause decision rule conflict,the algorithm adoptsthe idea of giving the imputation priority to decision attributes and introduces the concept of object completeness to improve it,and uses the improved ROUSTIDA algorithm for one-off preprocessing of data imputation on incomplete decision table,as well as employs attribute significancewhen in limited tolerance relation as the heuristic function to construct decision tree,so as to obtain the decision rule. Examples show that the method is effective,the generated decision rule is simple and has a higher accuracy.

关 键 词：不完备信息系统 对象完备度 规则提取 决策树 数据填补 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]