基于PCA-RBF神经网络的PPP项目风险智能评价研究

Intelligent Evaluation Model Based on PCA-RBF Neural Network Applied to Risk Assessment of PPP Projects

随着新一轮政府和社会资本合作(public—private—partnership,PPP)模式在基础设施领域的推广运用,对其风险评估也开始备受关注。鉴于PPP项目风险影响因素众多和传统评估方法过度依赖主观评价等问题,在构建PPP项目风险评价指标体系的基础上,通过将主成分分析(principal component analysis,PCA)技术降维并结合模糊综合评价结果,建立自适应的径向基神经网络(RBF)的智能风险评价模型,并以入库的浙江省发改委10个PPP项目为例进行实证检验,结果显示3个项目处于风险较小、6个项目处于风险不大、1个项目处于风险中等。PCARBF智能评价模型所训练预测的结果与模糊综合评价结论一致,且在一定程度上提高了计算效率、增强了评价的客观性,为PPP项目风险智能评估提供理论上的参考。

With new round promotion and application of PPP （public--private--partnership） model in infi-astructure, much attention has been given to its risk assessment. In view of the many influence factors to the PPP project risk and traditional methods relying too much on subjective evaluation, this paper, on the basis of building the PPP project risk evaluation index system, develops an adaptive RBF neural network intelligent risk evaluation model by PCA dimension reduction technology combined with fuzzy comprehensive evaluation results, taking for example those ten PPP projects put in storage by Zhejiang Provincial Development and Reform Commission for empirical test. Results show less risk for three out of them, little risk for six and medium risk for one. The results of training and prediction of PCA-RBF intelligent evaluation model is consistent with the fuzzy comprehensive evaluation conclusion, which improves the computational efficiency, and enhances the objectivity of the evaluation. It also provides theoretical reference for intelligent assessment of PPP project risks.