【作者】 周民志；

【导师】 周泓； 张在旭；

【作者基本信息】 中国石油大学 ， 石油工程管理， 2008， 博士

【摘要】 近50年来,世界石化企业发展方兴未艾,中小石化企业成为中国经济中不可忽视的一部分。为解决中小石化企业融资问题,规范经营行为,防范信贷风险,建立严格的信用评价体系势在必行。信用评价理论自19世纪中后叶起步以来,已建立了较为规范的指标评价体系和信用等级体系,也随着经济的发展进一步推进。信用评价方法也经历了从传统评价到现代统计技术评价,从定性评价到定量评价、定性与定量科学结合评价的发展历程,并逐步推进和完善。本文在系统总结传统评价方法、多变量评价方法、资本市场方法、衍生工具方法、信用风险集中评价模型的基础上,归纳了多因素多变量判别分析理论,并定义了基本概念,给出了基本数理解释,提出了多因素多变量判别分析的一般评价步骤。结合石化行业垄断加速、行业壁垒、资金密集等特点,定义了中小石化企业标准,根据信用风险要点分析了中小石化企业信用评价的要素指标,在创新传统5C评价方法的基础上对信用评价指标按基本素质、资本实力、营运能力、发展能力、形势环境进行了分类,运用专家调研法、变异系数法和相关系数法相结合的方法,确定了中小石化企业信用评价指标体系,建立了等级标准,为规范行业评级体系开辟了思路。针对企业信用综合评价和即期试点评价两种决策情景,提出了运用模糊评价法、层次分析法构造模型解决多因素多层次项下的中小石化企业信用综合评价,给出了构建步骤、计算和检验方法;将logit二项违约概率模型拓展为多元累积信用等级概率模型,并运用主成分分析法进行多变量降维处理,在最大限度保留基础数据信息含量的基础上提高了运算效率,奠定了科学定量计算信用等级的基础。运用实例验证了多因素多层次模糊综合评价方法和多变量logit累积概率模型,并对其一致性进行了检验比较;充分说明了两种模型应用的可行性和科学性。在信用评价的基础上,结合商业银行信贷风险评估的需求,通过样本测试和训练给出了商业银行对中小石化企业信贷风险评估的Logit二项模型;设计了BP神经网络算法,通过测试和训练,与Logit二项模型运算法进行了比较;实证表明运用两种模型进行信贷风险评估切实可行、科学有效。更多还原

【Abstract】 In the recent 50 years, the development of petrochemical enterprises all around the world have been in the ascendant, and small and medium-sized petrochemical enterprises in China are playing an important role in China’s economy. It is imperative to establish the strict credit evaluation system to settle the financing problems, to normalize business behaviors and to prevent credit risks in the small and medium-sized petrochemical enterprises. Since the credit evaluation theory started to be used in the late middle ages of 19th century, more standard index evaluation system and credit grading system have been set up, which has been carried forward along with the expanding economy. The method of credit evaluation has experienced the development changes from the traditional to modern statistic techniques, from qualitative to quantitative and in the scientific combination of qualitative and quantitative evaluation, and has been promoted and improved step by step. This essay put forward the multifactor and multivariate discriminant analysis theory according to the systematic review of the traditional, multivariate, capital market, financial derivative instrument, the model of centralized credit risk evaluation methods, defined the basic concept, explained the general logistics and demonstrated the general procedures for performing multifactor and multivariate discriminant analysis. In consideration of the feature of the rapid trade monopoly, trade barrier and capital-intensive in petrochemical industry, the standards for the small and medium-sized petrochemical enterprises are defined. The factors index of the credit evaluation on the small and medium-sized petrochemical enterprises are analyzed according to the essential points of credit risk. The credit evaluation index is classified into basic quality, capital strength, business management ability, development ability and the environment of the situation on the basis of innovation of the traditional 5C evaluation methods. The credit evaluation index system used by the small and medium-sized petrochemical enterprises is finalized, and the grading standards are constituted in combination with specialist research ,variation coefficient and correlation coefficient methods so that the thought of normalizing trade grading system is opened up. Under two decision-making circumstances of overall evaluation on the enterprise credit and immediate pilot evaluation, the apply of fuzzy evaluation and hierarchy evaluation methods are raised to build the model to deal with the overall credit evaluation on the small and medium-sized petrochemical enterprises in terms of multifactor and multilayer conditions as well as the design process and calculating and testing methods are given. The Logit PD (probability of default)model is developed into multivariate accumulation credit grade probability model, the application of principle component analysis method processed the multivariate dimensionality deduction, the efficiency of operation is enhanced as the utmost basic data information is kept, which establishes the basis to measure the credit grade by scientific quantitative calculation. The application of example has verified the multifactor and multilayer fuzzy overall evaluation method and multivariate logit accumulation probability model, and their consistency has been compared with and inspected, which the feasibility and scientificity of applying the two models is also fully illustrated. Based on the credit evaluation and in accordance with credit risk evaluation requirement of the commercial bank, the logit model is provided to the commercial bank for their credit risk evaluation on the small and medium-sized petrochemical enterprises after practice and testing of the sample model. BP Neural Network in the calculation is designed , through the test and practice and in comparison with logit fuzzy operational method, it is practical, scientific and effective to use these two models to implement credit risk evaluation.更多还原