基于数据挖掘技术的建筑企业信用评价

Credit Evaluation of Construction Companies Based on Data Mining

针对传统方法的不足,分析了应用数据挖掘技术的建筑企业信用评价方法.采用Logistic,决策树和神经网络算法,从250个建筑企业组成的学习样本中挖掘信用好或差的分类规则,从而建立了3个相应的信用评价模型.将所建立的模型用于评价检验样本中的46个建筑企业,采用混淆矩阵比较了各模型的评价表现.结果显示,Logistic,决策树和神经网络模型的评价准确率分别为87.0%,82.6%和82.6%,一致性结果的准确率达到91.7%,并且各模型在稳定性、敏感度等方面具有不同特点.研究表明,数据挖掘技术是一种有效而准确的建筑企业信用评价方法,此外,不同特点的数据挖掘模型为建筑业的信用评价提供了多种选择.

Because of the shortage of traditional methods, data mining was used to evaluate the credit of construction companies. Logistic, decision tree and neural network algorithms were employed in the learning sample with 250 construction companies to find the rules of classifying a construction company to good or bad credit, so three credit evaluating models were established with the rules. These models were used to evaluate 46 construction companies in the testing sample, and the performances of these models were compared by means of confusion matrix. The comparison indicates that the accuracy of Logistic, decision tree and neural network is 87.0%,82.6% and 82.6% respectively, and the accuracy of consistent judgment of the three models reaches 91.7%. Furthermore, the characteristics such as stability and sensitivity of each model were discussed. The results show that data mining is an effective and accurate method to evaluate the credit of construction companies, and the models with different characteristics provide alternative choices for the credit evaluation of construction companies.