基于逻辑回归评分卡的石油企业供应商风险模型研究  

Research on the Supplier Risk Model of Petroleum Enterprise Based on Logistic Regression Scorecard

作　　者：曹杰 张岩松 刘速 杨文军 高峰 刘增霞 CAO Jie;ZHANG Yansong;LIU Su;YANG Wenjun;GAO Feng;LIU Zengxia(Kunlun Digital Technology Co.,Ltd.)

机构地区：[1]昆仑数智科技有限责任公司

出　　处：《油气与新能源》2021年第5期51-57,共7页Petroleum and new energy

摘　　要：石油企业在进行采购选商时更多的依赖以往的采购经验和专家判断,为了进一步提升采购效率、节约采购成本,建立一套供应商风险评估模型有着重要意义。基于石油企业进项发票数据和供应商主数据提炼出基础属性、经营表征、发票表征等特征,通过数据预处理和信息值筛选出9个特征变量,运用逻辑回归模型拟合模型参数,构建逻辑回归评分卡。最终通过ROC曲线评估模型效果,基于Lift提升度验证评分卡在供应商风险评估上的适用性,成功将采购供应商通过不同风险等级分类管理。基于逻辑回归评分卡方法使采购管理者能够提升对供应商的风险管理,通过洞察供应商有关风险影响指标,辅助采购决策。Petroleum companies depend more on the previous procurement experience and expert judgment when purchasing and selecting suppliers.In order to improve procurement efficiency and save procurement cost,it is of great significance to build a set of supplier risk evaluation model.According to the input invoice data of petroleum companies and the master data of suppliers,the characteristics of basic attribute,management representation and invoice representation have been extracted.Besides,nine characteristic variables have been screened out through data pretreatment and information value,and logistic regression model has been used to fit model parameters to make logistic regression score card.Finally,the model effect has been evaluated by ROC curve.The adaptability of score card used for supplier risk evaluation has been verified based on Lift increase.And the purchasing suppliers have been classified and managed at different risk levels.By way of logic regression score card,procurement managers could improve their risk management ability when dealing with suppliers and assist procurement decision-making by gaining insights into the supplier-related risk impact indicators.

关 键 词：逻辑回归 评分卡 供应商 风险模型 石油企业 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构] TE-9[自动化与计算机技术—计算机科学与技术]