一种基于机器学习的属性缺失值模糊填补方法

A Fuzzy Imputation Method for Missing Attribute Values Based on Machine Learning

针对不完全信息多属性决策问题中属性值缺失的情况,为使缺失值的填补更加客观,填补后数据集整体尽量保持填补前的分布,且不丢失已有信息,提出了一种基于机器学习的属性缺失值模糊填补方法。该方法通过寻找不需填补的属性相似的记录,在这些记录中发现需填补属性的可能取值及其概率,按照各取值的概率为缺失值分配相应的取值。该方法的基本思想对于离散型和连续型的数据集均适用。

To missing attribute values of Multiple Attribute Decision Making（MADM） problem with incomplete information, this paper brings forward a fuzzy imputation method for missing attribute values based on machine learning, for complete missing values more objectively and making the whole dataset keep distribution as similar as possible as before, and not missing obtained information. This method searches the records which attributes are alike and need not to be completed, and finds out the possible value-taking and probability which need complete attribute, and assigns corresponding value in accordance with respective valuetaking probability. The basic idea of the method can be applied to both discrete and continuous dataset.