基于ROC曲线的建设工程招投标异常低价检测模型研究

On the Abnormal Low Price Detection Model of Construction Project Bidding Based on ROC Curve

为判别招投标工作中的异常低价,文章采用参数和非参数估计方法拟合概率分布,以累积分布函数(CDF)为评价指标,通过Monte Carlo模拟和ROC曲线分析得到最优阈值,以此判别异常低价投标,并划分风险等级。结果表明,对少部分服从正态分布的项目投标报价,采用最大似然估计(ML)法和非参数核密度估计(KDE)法拟合结果效果良好,对不服从正态分布的项目投标报价,使用KDE法拟合效果较好;采用CDF作为评价指标可有效界定低价区间,评价中标价格的风险等级。最后以某省6条高速公路项目投标数据为例,验证该方法的适用性与有效性。

In order to identify the abnormal low price in bidding,parametric and non-parametric estima⁃tion methods is used in this paper to fit the probability distribution.Taking cumulative distribution function(CDF)as the evaluation index,the optimal threshold is obtained through Monte Carlo simulation and ROC curve analyses,so as to identify the abnormal low price bidding and divide the risk level.The results show that the maximum likelihood estimation(ML)and nonparametric kernel density estimation(KDE)method can be used to fit the bidding price of a small number of items conforming to the normal distribution.The KDE method has a better fitting effect for the project bids that do not obey the normal distribution.Using CDF as the evaluation index can effectively define the low price range and evaluate the risk level of the bid price.Finally,the application and effectiveness of the proposed method are verified by taking the bidding data of 6 expressway projects as an example.