基于支持向量回归机的中国物价波动影响因素探究

发布时间：2018-01-30 22:25

本文关键词： 支持向量回归机物价波动特征选择小波核　出处：《浙江工业大学》2014年硕士论文　论文类型：学位论文

【摘要】：物价不仅反映市场的冷热程度,且为实现资源的合理配置扮演着指示器的功能。保持平稳的物价水平是货币政策当局关注的焦点。进入21世纪以来,我国经历了几次明显的物价波动,其影响力度逐渐加强。在宏观经济形势日益复杂的情况下,分析和掌握物价波动的成因以及有效预测物价波动趋势对我国未来经济发展、人民生活水平提高意义重大。支持向量回归机(Support Vector Regression,简称SVR)是近年来发展起来的数据挖掘新技术,有强大的理论基础,已成功应用于多个领域。本文引入此方法对物价波动的成因以及未来走势进行了如下探讨： 1.结合“结构风险最小化”原则和“最大间隔”思想,构造具有稀疏性的特征选择模型(L1-ε-TSVR)研究我国汇改前后物价波动的影响因素。数值试验验证了模型的有效性,通过与ε-TSVR模型以及OLS方法的比较,突出了该模型的优势。 2.在L1-ε-TSVR模型中引入小波核变换构建L1-ε-WTSVR模型预测物价波动趋势。模型中采用小波核比高斯核预测效果更佳,且优于VAR方法。该模型拓宽了SVR在中国金融领域的应用,为当局制定合理的货币政策工具维持物价稳定以及宏观经济的健康发展做出绵薄之力。本文在SVR的理论框架下,实证分析得到以下结论：1)汇改前中国物价主要受成本型因素推动,其中原材料、燃料、动力购进价格指数成为决定性因素,而利率、汇率等货币因素几乎不构成影响；2)汇改之后,物价波动的成因更具有多面性和复杂性,除了受成本型因素推动外,需求和结构型因素的影响力度也逐渐加强。此阶段利率成为影响物价波动最重要力量,汇率的影响强度亦显著增强；3)2013年上半年中国物价上涨压力明显较2012年有所增加,但走势基本处于温和可控状态；4)新时期若要有效控制通胀,保持物价稳定,需要综合运用多种手段,形成增加有效供给、保障市场需求、稳健货币政策、稳定人民币汇率预期的组合策略,从而有效提高政策控制力。
[Abstract]:Prices not only reflect the cold and hot degree of the market, but also act as indicators for the rational allocation of resources. Maintaining a stable price level is the focus of monetary policy authorities. Since 21th century. Our country has experienced several obvious price fluctuation, its influence strength strengthens gradually. In the macroeconomic situation increasingly complex situation. It is of great significance to analyze and grasp the causes of price fluctuation and to effectively predict the trend of price fluctuation for the future economic development of our country and the improvement of people's living standard. Support Vector regression (SVR) is a new data mining technology developed in recent years, which has a strong theoretical foundation. It has been successfully applied in many fields. In this paper, the causes and future trend of price fluctuation are discussed as follows: 1. Combine the principle of "structural risk minimization" with the idea of "maximum spacing". A sparse feature selection model (L1- 蔚 -TSVR) is constructed to study the influencing factors of the price fluctuation before and after the exchange rate reform in China. The validity of the model is verified by numerical experiments. Compared with 蔚 -TSVR model and OLS method, the advantages of this model are highlighted. 2. Using wavelet kernel transform to predict price fluctuation trend in L1- 蔚 -TSVR model. Wavelet kernel is better than Gao Si kernel in predicting price fluctuation trend. The model extends the application of SVR in China's financial field and makes a small contribution to the establishment of reasonable monetary policy tools to maintain price stability and the healthy development of macroeconomic. In this paper, under the theoretical framework of SVR, empirical analysis of the following conclusions: 1) before the exchange rate reform, China's prices are mainly driven by cost factors, in which raw materials, fuel, power purchase price index is the decisive factor. But the interest rate, exchange rate and other monetary factors have almost no effect; 2) after the exchange rate reform, the causes of price fluctuation are more multifaceted and complicated, except by the cost type factors. The influence of demand and structural factors is gradually strengthened. At this stage, interest rate becomes the most important force affecting price fluctuation, and the influence intensity of exchange rate also increases significantly. 3) in the first half of 2013, the pressure of price increase in China was obviously higher than that in 2012, but the trend was basically in a moderate and controllable state. 4) in order to effectively control inflation and maintain price stability in the new period, it is necessary to comprehensively use a variety of means to form a combination strategy of increasing effective supply, ensuring market demand, sound monetary policy, and stabilizing the expectation of RMB exchange rate. So as to effectively improve policy control.
【学位授予单位】：浙江工业大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：F224;F726

【参考文献】