代价敏感的缺失数据有序填充算法  

Cost-sensitive Missing Data Imputing Algorithm with Ordering

作　　者：苏毅娟[1] 钟智[1] 

机构地区：[1]广西师范学院计算机与信息工程学院,南宁530023

出　　处：《计算机工程》2009年第17期92-93,96,共3页Computer Engineering

基　　金：国家"973"计划基金资助项目(200807MS176);广西教育厅科学研究基金资助项目(2008CB317100)

摘　　要：缺失数据填充效果会对学习算法和挖掘算法的后续处理过程产生影响。针对代价敏感决策树方法没有同时考虑填充顺序和填充代价的问题,提出一种有序填充缺失数据的算法,综合考虑经济因素和建立填充器所需的有效信息。实验结果表明其预测准确率和分类准确率高于现有算法。Missing data imputing effect affects the following processes of the learning algorithms and mining algorithms. Cost-sensitive decision tree method does not consider the imputing order and imputing cost at the same time. Aiming at this problem, this paper proposes a new algorithm to impute missing data with ordering. This algorithm considers the economic factor and effective information for imputing machine establishment synthetically. Experimental results demonstrate that this algorithm has high prediction accuracy and classification accuracy than existing algorithms.

关 键 词：代价敏感学习 缺失数据填充 填充顺序 

分 类 号：TP331[自动化与计算机技术—计算机系统结构]