摘要：为解决高速公路项目投资管控中风险隐患隐蔽性强、因果关系复杂的问题，提出一种基于审计文本挖掘与贝叶斯网络的风险评估模型。首先，利用文本挖掘技术对F省近一年高速公路项目审计报告进行结构化处理，构建了4类共36项风险因素集。其次，通过Apriori关联规则挖掘与专家知识结合，确立风险传导路径并构建贝叶斯网络拓扑结构。最后，以F省A项目为实证对象，综合运用正向推理、逆向诊断及敏感性分析进行量化评估。研究结果表明：A项目投资管控风险呈现出复杂网络结构特征；征迁安置组织不力与项目资源配置滞后是具有高敏感度的根源性风险，以清单勘误代替设计变更是关键隐蔽性合规风险。该方法实现了高速公路项目投资管控从“事后审计”向“事前预警”的转变。

关键词：高速公路项目；投资管控风险；贝叶斯网络；风险评估；审计报告

Abstract：To address the issues of high concealment and complex causality regarding investment control risks in expressway projects，this paper proposes a risk assessment model based on audit text mining and Bayesian Networks. First，text mining technology is employed to structure the audit reports of expressway projects in F Province from the past year，and a set of 36 risk factors across four categories is constructed. Second，by combining Apriori association rule mining with expert knowledge，risk transmission paths are identified and the Bayesian Network topology is built. Finally，taking Project A in F Province as a case study，a quantitative assessment is conducted through forward reasoning，backward diagnosis，and sensitivity analysis. The results indicate that the investment control risk of Project A exhibited characteristics of a complex network structure. Specifically，ineffective organization of relocation and resettlement and delayed allocation of project resources are identified as high-sensitivity root risks，while substituting design changes with Bill of Quantities corrections is a critical hidden compliance risk. This method realizes the transition of expressway project investment control from "post-audit" to "early warning".

Keywords：highway project；investment control risk；Bayesian network；risk assessment；audit report

参考文献

[1] 李晓东.浅谈高速公路建设项目竣工决算审计程序和方法[J].公路，2014（10）：208-211.

[2] 石水平，吴文强，何敏燕，等.研究型审计在高速公路建设项目中的应用与实践——以G项目成本控制审计为例[J].会计之友，2024（11）：41-46.

[3] JOBY P J，KORRA J. Accessing Accurate Documents by Mining Auxiliary Document Information[C]//2015 Second International Conference on Advances in Computing and Communication Engineering，2015：634-638.

[4] AYDOAN M，KARCI A. Improving the accuracy using pre-trained word embeddings on deep neural networks for Turkish text classification[J]. Physica A：Statistical Mechanics and its Applications，2020（541）：123288.

[5] 姜建武，王博.高维数据组合关联关系挖掘方法[J].科学技术与工程，2023（4）：1615-1624.

[6] 马德仲，周真.基于模糊概率的多状态贝叶斯网络可靠性分析[J].系统工程与电子技术，2010（12）：2607-2611.

[7] 林志军，李敏，贺珊，等.基于博弈论—贝叶斯网络的煤矿瓦斯爆炸风险评估[J].煤炭学报，2024（8）：3484-3497.

[8] 黄国忠，姜莉文，谢志利，等.基于模糊贝叶斯网络的电动平衡车失速事故发生可能性研究[J].安全与环境学报，2018（6）：2081-2085.

建筑经济，2026（3）：66-73