基于非负矩阵分解的缺失数据插补算法研究

摘要

Supply chain performance depends on the real supply chain performance data. Because of lots of node enterprises, the management consulting commonly used questionnaires, interviews and other research methods cannot meet the actual needs, the data quality of supply chain performance diagnostic a-nalysis is difficult to guarantee. Focus on the interpolation of missing values in the sample data, mainly for the general lack of pattern in the missing data problem, especially for supply chain performance data missing. Based on the missing data processing technology, decomposing matrix theory, and data mining technology, effective imputation algorithm has been given, which based on the decomposition of the nonnegative matrices factorization. The effectiveness of the algorithm is demonstrated by the numerical experiments. The algorithm can be more efficient, accurate, low-cost to get performance data on each node in the supply chain enterprises.%供应链绩效分析依赖于真实的供应链绩效数据.对于供应链绩效诊断分析问题,由于节点企业众多,管理咨询常用的发放问卷、访谈等调研方式难以满足实际需求,数据质量也难以保证.重点探讨样本数据中的缺失值插补问题,主要面向缺失数据问题中的一般缺失模式,特别针对供应链绩效数据缺失问题.基于缺失数据处理技术及矩阵分解理论,给出有效的插补算法:基于非负矩阵分解的插补算法,通过数值实验证明了算法的有效性.该算法可以更为有效、准确、低成本地获得供应链上各节点企业的绩效数据.