非平衡样本集下公私合作(PPP)可融资性评价——基于改进边界样本自适应算法

Financability Evaluation of Public-Private Partnership Under Non-equilibrium Sample Set:Based on Borderline-SMOTE Bagging Algorithm

可融资性难问题持续制约PPP健康发展,社会资本往往从项目本身和地方政府两个维度评价项目的可融资性。通过综合比较主流样本合成算法在合成样本的精细度以及分类器算法对非平衡样本集少数类样本的识别能力,针对我国财政部政府和社会资本合作中心库中PPP案例样本数据存在非平衡性及高噪声异质性问题,提出Borderline-SMOTE Bagging算法,按照不同领域对其中4组项目进行可融资性评价。结果表明:基于数据挖掘算法对PPP可融资性进行评价具备可行性;Borderline-SMOTE Bagging算法具备良好的样本分类能力和优秀的泛化能力,能有效降低因合成样本形成的噪音所带来的负面影响,且具备良好的少数类样本识别能力。最后结合实证过程遇到的问题,对未来PPP数据化发展,提出政府部门应增强PPP项目数据收集能力并逐步实现数据开放共享,借助大数据技术提升PPP项目管理效率和精准度等建议。

The difficulty of financing continues to restrict the healthy development of PPP,and social capital often evaluates the financability of projects from two dimensions of the project itself and the local government.By comprehensively comparing with main stream sample synthesis algorithm in the fineness of synthesis samples,and the classifier algorithm to identify few classes of non-equilibrium sample sets,aimingat the problems of non-equilibrium and high noise heterogeneity in sample data of PPP cases in China Public Private Partnership Center(CPPPC),this paper proposes the Borderline-SMOTE Bagging algorithm to evaluate the financing of 4 groups PPP projects in different fields.The results show that it is feasible to evaluate the PPP financability based on data mining algorithms;the Borderline-SMOTE Bagging algorithm has good sample classification capabilities and excellent generalization capabilities,which can effectively reduce the negative impact caused by the noise formed by synthetic samples,and has good minority sample identification ability.Finally,combined with the problems encountered in the empirical process,for the future data-based development of PPP,the paper suggests that government departments should enhance the data collection ability of PPP project and gradually realize data opening and sharing,and improve the efficiency and accuracy of PPP project management with the help of big data technology.