数据挖掘中基于ICA的缺失数据值的估计

本文简单介绍了数据挖掘中缺失数据的研究现状及ICA的特点与发展前景,提出了基于ICA的缺失数据估计模型——ICA-MDH模型。该模型研究了数据之间存在相关关系且为非高斯分布时缺失数据的处理方法,该方法能充分利用已知数据记录中的已知信息,且具有较好的通用性。实验通过对一些不完整经济数据进行了处理。结果表明,本文提出的缺失数据估计方法的精度明显优于平均值法和PCAs法,从而验证了本文所提模型的正确性与合理性。

Missing Value Estimation Based on ICA in Data Mining

This paper introduces the study state of missing data as well as ICA's characteristics and foreground in brief,proposes a model,which is named as ICA-MDH model in this paper,based on ICA which is used abroad to dis- pose missing data under the circumstances of that data remain correlative and non-Gaussian distribution.This model takes full use of the known information of the given data to estimate missing data.By dealing with some economic data, the experiment verifies the results that ICA method is much better than PCAs and mean method,validating the correct- ness and reasonableness for the proposed model in this paper.