非平衡样本集下的PPP可融资性评价研究——基于Borderline-SMOTE Bagging方法

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PPP项目可融资性。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识别能力，针对CPPPC库中PPP案例样本数据非平衡性及高噪声异质性，提出Borderline-SMOTE Bagging算法，对四组PPP项目进行可融资性评价。研究结果表明：基于数据挖掘算法对PPP可融资性进行评价具备可行性；针对PPP非平衡数据集问题，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。

A Research on the Financing Evaluation of PPP Under Imbalanced Dataset:Based on Borderline-SMOTE Bagging Algorithm

Financing difficulties continue to restrict the healthy development of PPP. Social capital often evaluates the financability of PPP projects from the two dimensions of project itself and local government. By comprehensively comparing the fineness of synthetic samples of mainstream sample synthesis algorithms and the ability of classifier algorithms to identify minority samples in imbalanced dataset, the Borderline-SMOTE Bagging algorithm is proposed to evaluate the financability of PPP projects in CPPPC, four groups of PPP projects are evaluated for their financing ability.The research results show that it is feasible to evaluate the financing of PPP based on data mining algorithms; Aiming at the problem of PPP imbalanced dataset,the Borderline-SMOTE Bagging algorithm has good sample classification capabilities and excellent generalization capabilities with imbalanced dataset of PPP in the field of technological innovation. It can reduce the negative impact of noise caused by synthetic samples effectively, and has good minority sample recognition capabilities.