Missing Data estimation with a bi-dimensional adaptive weighted method for power grid data

2nd International Symposium on Application of Materials Science and Energy Materials

材料科学;能源学

Wang, Xi^1 ; Li, Mingwei^1 ; Zhou, Jianjia^1 ; Deng, Yan^1 ; Zhang, Qingqing^2

Sichuan Electric Power Company, State Grid Corporation of China, China^1

School of Automation Engineering, University of Electronic Science and Technology of China, China^2

关键词: Development trends;    Estimation errors;    Imputation methods;    K-nearest neighbors;    Missing data estimation;    Missing values;    Two directions;    Weighted method;   

Others  :  https://iopscience.iop.org/article/10.1088/1757-899X/490/4/042025/pdf
DOI  :  10.1088/1757-899X/490/4/042025

学科分类：材料科学（综合）

来源: IOP

【 摘 要 】

The power grid data often contain missing values, which will seriously affect the analysis of the power grid development trend. To estimate missing values, the k-nearest neighbor and linear regression imputation method have been proved to be effective. These methods only consider the correlation between the samples or just consider the correlation between the features. However, we do not know which correlation of the two dimensions is more important in some special missing values. So this paper proposes a bi-dimensional adaptive weighted imputation method which considers the correlation of the two directions simultaneously. The experimental results show that the bi-dimensional adaptive weighted imputation method can reduce the estimation error.