高新技术企业资格维护风险预警模型研究——以河北省高新技术企业为例  

Research on the risk early warning model of high-tech enterprise qualification maintenance——Take Hebei Province's high-tech enterprises as an example

作　　者：王云丽[1] 韩珍珍 杨文焕 成彬[1] WANG Yun-li;HAN Zhen-zhen;YANG Wen-huan;CHENG Bin(Hebei Authentication Technology Engineering Research Center, Institute of Applied Mathematics, Hebei Academy of Sciences, Shijiazhuang Hebei 050081, China)

机构地区：[1]河北省科学院应用数学研究所,河北省信息安全认证技术创新中心,河北石家庄050081

出　　处：《河北省科学院学报》2022年第3期59-67,共9页Journal of The Hebei Academy of Sciences

基　　金：河北省创新能力提升计划项目(21550301D);河北省科学院两院合作项目(21607)。

摘　　要：针对高新技术企业认定后的资格维护情况,分析了高新技术企业运行发展的公开特征指标,利用逻辑回归建立了高新技术企业资格维护风险预警模型。以河北省高新技术企业为例,从企业基本信息、经营状况、经营风险和知识产权等四个方面采集80余个特征维度的数据。将高新技术企业资格“被取消后再次获取”作为预警模型的标签,采用上采样方式平衡样本,运用嵌入法降维,进一步使用IV值进行筛选,得到企业和标签之间有明显贡献度的11个特征,构建逻辑回归预警模型,并根据WOE值制作评分卡。In view of the qualification maintenance situation of high-tech enterprises after the identification,the public characteristic indicators of the operation and development of high-tech enterprises are analyzed in this paper,and logistic regression is used to establish a high-tech enterprises qualification maintenance risk early warning model.Following the obtaining of the high-tech enterprises list in Hebei Province,data of more than 80 characteristic dimensions is collected from four aspects:basic information,operating conditions,operating risks and intellectual property rights of the enterprise.With the high-tech enterprise qualification"re-acquire after being canceled"treated as the label of the early warning model,up-sampling balancing the samples,the embedding method reducing the dimensionality,and the IV value used for screening,11 features with obvious contribution between the enterprises and labels are obtained,and a logistic regression early warning model is constructed,and a score card based on the WOE value is made.

关 键 词：高新技术企业 资格维护 逻辑回归 特征分析 

分 类 号：TP391[自动化与计算机技术—计算机应用技术] F276[自动化与计算机技术—计算机科学与技术]