基于GM-BP神经网络的地方政府债务风险预警系统研究——以2015年重庆市38区县样本数据为例  

Local Government Debt Risk Warning System Based on GM-BP Neural Network——A Case Study of 2015 Sample Data from 38 Chongqing Districts and Counties

作　　者：马恩涛[1] 吕函枰 MA Entao LV Hanping(School of Finance and Taxation, Shandong University of Finance and Economics, Jinan 250014, China)

机构地区：[1]山东财经大学财政税务学院,山东济南250014

出　　处：《山东财经大学学报》2017年第2期96-106,共11页Journal of Shandong University of Finance and Economics

基　　金：国家社会科学基金项目"我国地方政府融资平台债务控制及其风险防范研究"(13BJY164);山东省自然基金重点项目"山东省政府性债务控制及风险预警研究"(ZR2015GZ001);济南市社科规划基金重点项目"济南市政府与社会资本合作中的问题与对策研究"(JNSK16B04)

摘　　要：以2015年重庆市38个区县的债务数据作为研究样本,利用灰色关联方法(GM)与BP神经网络两种理论在非线性处理方面的优势,构建了基于GM-BP神经网络的地方政府债务风险预警系统,并运用该预警系统对重庆市各区县债务风险进行了实证分析。结果表明:2015年重庆市33个区县债务风险处于绿色可控区,4个区县(大渡口区、开县、南川区、潼南区)处于橙色预警区,1个区县(城口县)债务风险处于红色风险区,重庆市地方政府债务风险总体可控;并且,与未经约简的BP神经网络预警系统相比,GM-BP神经网络预警系统的训练时间更短,预警准确性更高,在结合预警地区的实际情况做出微调后,其更具有一定的普适性。With the 2015 debt data from 38 Chongqing districts and counties as sample and via the nonlinear processing advantages of grey correlation method and BP neural network, this paper constructs a GM-BP neural net- work-based local government debt risk warning system and empirically analyzes the debt risks of all Chongqing dis- tricts and counties by adopting this warning system. The results show that in 2015 the debt risks of 38 Chongqing dis- tricts and counties remain in the Green Zone with 4 districts and counties （ Dadukou District, Kaixian County, Nanchuan District and Tongnan District） in the orange warning area and 1 county （Chengkou County） in the red risk area while Chongqing local government debt risks are overall controllable, and that compared with the unreduced BP neural network warning system, the GM-BP neural network debt risk warning system needs a shorter training time, has higher early warning accuracy, and presents a certain degree of universality if fine-tuned according to the actual situation of early warning areas.

关 键 词：灰色关联方法 BP神经网络 地方政府债务风险 预警系统 

分 类 号：F810.7[经济管理—财政学]