【作者】 鞠可一；

【导师】 周德群；

【作者基本信息】 南京航空航天大学 ， 管理科学与工程， 2011， 博士

【摘要】 石油价格系统是一个典型的复杂系统。石油资源天然的稀缺性及其分布的不均衡性,使得石油价格除了会受到供求关系的影响外,还易受诸如经济形势、国际关系、突发事件、投机行为等诸多其他因素的影响,任何环节的微小变化都有可能造成石油价格的剧烈波动。与此同时,石油价格波动对经济产生的影响也越来越明显。早在1983年,著名经济学家Hamilton就指出,从第二次世界大战到1983年,全球7/8的经济萧条都伴随着高油价的出现。不断增大的供需缺口和日益上涨的石油价格,已经严重威胁到了我国的石油价格安全和经济安全,石油价格波动对经济的影响已经成为了一个不可回避的现实问题。预警是保障石油价格安全和经济稳定发展的有效手段,通过开展石油价格波动对经济影响的预警研究,可以实时、有效地对石油价格及经济运行状况进行监控,有针对性地对油价及经济运行过程中的不稳定因素进行干预和微调,从而减少石油价格波动对国民经济造成的冲击。然而大量的实证结果表明,现有的危机预警理论不足以探求国际石油价格波动的驱动机制,无法深入挖掘石油价格波动对经济影响的表现特征,因而也就无法为石油价格波动对经济可能造成的影响提供有效的预警信号。本文的研究建立在全新的预警理论基础之上,将数据挖掘技术和人工智能方法融入预警理论中,构建预警知识库系统。通过理论推导和系统分析,对石油价格波动及其对经济的影响展开实证分析和仿真研究。论文的主要工作和研究创新如下:(1)选用1986~2009年WTI周平均价格作为研究对象,探寻隐藏在石油价格波动过程中的周期性特征,为预警知识库系统的构建提供周期选择的依据。研究发现,石油价格波动的最小周期为4周(1个月)左右,且大规模石油危机的爆发周期约为12年。这一结论也可由历次石油危机的时间得到验证。已有的四次石油危机分别发生在:1973,1978,1990,2002年,除了前两次的石油危机间隔周期较短(仅为5年)外,其余石油危机爆发的周期均为12年。(2)采用AHHDOD算法,将石油价格波动对经济的影响以经济的“离群”特征来描述,用离群点的数量来衡量经济对石油价格波动的敏感程度。借助国际比较,探讨不同国家经济与石油价格之间的联动波动关系。结果表明,中国是受油价波动影响最为明显的国家,其次是印度,接着是日本和美国。石油价格波动对我国的GDP、CPI、进出口差额这三大领域均有显著影响,其中又以进出口领域受影响最大。(3)利用事件分析法,对油价——经济的关联关系进行了定量分析,回答了“石油价格在多大程度上对经济造成了影响”的问题。并通过国际比较,得出了在油价上涨和下跌阶段,经济波动趋势和中国最为接近的国家。结论指出,石油价格上涨和下跌阶段对我国经济产生的影响是非对称的,油价上涨对经济的影响远大于油价下跌的影响。当石油价格上涨时,与我国经济波动趋势最为接近的国家是印度,而当石油价格下跌时,经济波动趋势与我国最为接近的是欧盟地区。(4)将本体融入案例推理过程,设计混合概念格(Multi Galois Lattice,MGL)案例匹配算法,构建石油价格波动对经济影响的预警知识库系统。借助本体较好的知识共享能力,案例推理的自学习、自适应的特征和混合概念格的方法,优化案例的表示过程,提高案例检索的查全率和查准率。仿真结果显示,一方面,当输入检索条件较为模糊,或是在案例库中没有明确对应项时,案例检索系统能够通过本体中内含的语义关系检索出与查询条件在语义上相似的案例。另一方面,采用本文设计的MGL匹配算法检索出来的结果,其相似度要高于传统匹配算法的相似度,这使得MGL算法的有效性得到了进一步的验证。(5)对构建的石油价格波动对经济影响的预警知识库进行案例研究,以验证预警知识库的有效性。首先,分别对经济危机、自然灾害引发的石油价格波动过程对经济可能带来的影响进行预警研究。预警结果显示,知识库给出的预警信息基本符合经济的实际走势,从而证明了本文构建的石油价格波动预警知识库系统是可行的。此外,通过对混合概念格匹配算法与传统案例匹配算法的检索结果进行对比,验证了MGL检索算法在提高案例检索查全率和查准率时的有效性。最后,以人民币升值为研究背景,对人民币升值后国际石油价格波动情况以及对中国经济可能带来的影响进行预警分析,预警结果指出,一旦人民币升值,首先受到影响的是我国的进出口领域,而扩大内需的方法既能够转移过剩的产能,同时又能让普通消费者也享受人民币升值带来的好处。另外,人民币升值可能引发大量境外热钱的涌入,从而引发通货膨胀,应当适当提高外币兑换人民币的门槛,以防止投机资金对我国经济带来的冲击。更多还原

【Abstract】 Oil price market is a typical complex system. High consumption dependence, scarcity and uneven distribution made oil price be easily influenced by many factors, such as economic situations, international relations, emergent affairs, financial speculation, as well as fundamental factors like supply and demand. Even minor mutual interactions among these factors can induce unexpected outcomes, which will cause a severe shock on oil price. Meanwhile, the macroeconomic effects caused by oil price shocks become more and more prominent. Hamilton had pointed out that seven out of eight economic depressions were accompanied with high oil price since World War II to 1983. Growing gap between supply and demand as well as rising oil price has threatened China’s oil price security and macroeconomic security. The macroeconomic effects caused by oil price shock have become an unavoidable practical problem.Pre-warning is an effective method to protect oil price security and economic stability. However, a large number of empirical results show that, current crisis pre-warning theory can not fully explore the drive mechanism of international oil price shocks, and it can not thoroughly discover the performance features of the effects on macroeconomy of oil price shocks. In that case the current pre-warning system is not applicable for the analysis of macroeconomic impacts caused by oil price shocks.The study of this paper is built on the basis of a new pre-warning theory--data mining (DM) technology and artificial intelligence (AI) theory are added to construct a pre-warning knowledge base to find the relationship between oil price shock and macroeconomy. The main task of this paper is to give a rigorous theoretical analysis, and perform an applied research on the impacts of oil price shocks on macroeconomy. With the real-time monitor of oil price and macroeconomic situation, we can intervene and adjust the unstable factors of oil price shocks and macroeconomy, in order to minimize the negative effects on macroeconomy caused by oil price shocks. The main innovations and researches are as follows:(1) Weekly WTI oil prices are used for exploring the hidden periodicity features of oil price shocks. Results show that, the minimum period of one oil price shock is about four weeks (nearly one month), which is the periodic basis for pre-warning knowledge system. In addition, the period of large-scale oil crisis is about 12 years. Apart from the interval of 1973 and 1978 crisis is 5 years, the periods of the other two crises are all 12 years (1978, 1990, and 2002).(2) With AHHDOD algorithm, the macroeconomic impacts of oil price shocks will be depicted as the“Outlier”features of macroeconomic data sets, and the number of outliers is used to measure the sensitivity of macroeconomy to oil price shocks. Through the international comparison, we can explore the different sensitivity of different countries to the oil price shocks. Results show that, China is the most subject one to the oil price shock, followed by India, Japan, and America. Oil price shock affected China’s GDP, CPI, and gross imports and exports, among which the gross imports and exports are the most affected area.(3) Quantify the relationship between oil price shock and macroeconomy with event study method to answer the question: to what extent will oil price shock affect macroeconomy?”Simultaneously, select the countries which economic trends are similar with China when oil price rising and dropping. It can be concluded that, oil price rising and dropping caused asymmetrical impacts to China’s macroeconomy. Rising oil price shocks produce more impacts than dropping ones. On the other hand, India is the country whose macroeconomy fluctuate similar as China when oil price rising, while EU is the region whose macroeconomy fluctuate similar as China when oil price dropping.(4) Ontology is added into case-based reasoning for constructing the pre-warning knowledge base for the relationship between macroeconomy and oil price shocks. The knowledge sharing capability of ontology and the self-learning and self-adaptive capability of case-based reasoning are feasible for knowledge base construction. When type in“Storm disaster”, the case retrieve system can find out the cases of“Hurricane Ivan hit Gulf of Mexico”and“inclement weather reduced the oil production in North Sea”, all of which are the individuals of the same class“environment”. Additionally, we introduced the Multi-Galois Lattice (MGL) to enhance the recall and precision ratios during case retrieving. Results show that, the similarities of the cases retrieved by MGL are higher than those which are retrieved by the traditional method, which illustrates that the whole operation of MGL is vivid. (5) Taking case study on the pre-warning knowledge base of the macroeconomic impacts of the oil price shocks, to verify its effectiveness. Firstly, the paper makes some early warning for the possible macroeconomic impacts of oil price shocks caused by economic crisis and natural disasters. Results show that, the early warning signals are basically consistent with the actual trend of the macroeconomy, which shows the feasibility of the pre-warning knowledge base. Additionally, MGL retrieve algorithm is compared with the traditional retrieve algorithm called Interval retrieve algorithm, in order to prove that MGL is prior than Interval retrieve algorithm in enhancing the recall and precision ratios during case retrieving. At last, we take appreciation of RMB as a background to warn whether the oil price shock will affect China’s macroeconomy. The results indicate that, the appreciation of RMB will first and for most affect China’s international trade. The exports will decrease. It is the best way to expand domestic demand, which can not only transfer the excess produce capacity, but also allowing consumers enjoy the benefits of appreciation. Further more, RMB appreciation will attract a lot of foreign currency coming into China, and causing inflation. We have to increase the threshold of foreign currency exchange with RMB, in order to prevent speculative capital impacts on our macroeconomy.更多还原