基于神经网络的大型国际工程财务风险控制评价

Financial Risk Control Evaluation in International Megaprojects Based on Back Propagation Neural Network

摘要:

建立了一个由7个一级指标和32个二级指标构成的系统评价体系.采用模糊层次分析法(FAHP)并结合专家问卷数据得到初始指标权重，进而生成30组前馈神经网络（BPNN）训练样本和5组校验样本.基于训练样本和前馈神经网络构建出大型国际工程项目财务风险控制体系的非线性映射关系模型，并通过样本校验.最后，以某大型央企在沙特的总承包项目为例，进行了评价体系的实际应用和案例分析.

Abstract:

An evaluation framework consisting of 32 evaluation indicators grouped under 7 categories was built. The weights of the indicators were obtained through an expert questionnaire survey and were then computed using fuzzy analytic hierarchy process(FAHP), which generated 30 groups of back propagation neural network(BPNN) training samples and 5 sets of calibration samples. Based on the results, a nonlinear mapping relation model of the financial risk control of international megaprojects was constructed. Finally, an EPC(engineering, procurement, construction) project in Saudi Arabia was used as a case to illustrate the evaluation system.