基于神经网络的我国通货膨胀预测研究

发布时间：2018-07-10 05:15

本文选题：神经网络 + 遗传算法　；参考：《湖南大学》2012年硕士论文

【摘要】：作为判断经济是否稳定的先期指标,通货膨胀率直接或间接影响着工资、利率、汇率等宏观经济指标。从上世纪90年代开始,将通货膨胀率控制在合理水平,以促进经济的平稳发展逐渐成为各国中央银行货币政策的中心目标。在此背景下,通货膨胀预测作为一种新的货币政策中介目标,在提高货币政策有效性,以稳定物价方面发挥着越来越重要的作用,尤其是在实行通货膨胀目标制的国家。目前,虽然通货膨胀预测方法众多,但皆存在着不足之处,而且央行的决策也不仅仅参考一种模型结果,因此,构建一种新的预测方法,为央行决策时提供对比参考便具有理论与现实意义。本文遵循从理论阐述、实证分析到对策建议的思路展开。神经网络作为新兴的预测方法,以其非线性、基于数据驱动和稳定性高等优点在国外得到了广泛应用,国内虽亦有众多研究,但在通货膨胀预测方面却较少。因此本文在借鉴国内外相关研究成果的基础上,采用定性分析与定量分析相结合的方法,从理论和实证角度对神经网络在我国通货膨胀预测方面的适用性进行了研究。文章首先简要介绍了神经网络和通货膨胀预测的基础理论,对相关概念进行了说明。然后罗列了在通货膨胀预测方面经验丰富的部分国家的工作流程以及特点,对比分析了我国在此方面存在的不足。并利用我国通货膨胀部分影响因素2005年3月到2011年12月的月度数据构建了BP神经网络模型进行短期预测。实证结果表明经优化后的网络能根据已有数据对未来至少6个月的CPI进行良好预测,同时未经优化的网络也给出了较为合理的预测结果。文章最后就如何提高央行通货膨胀预测的准确性提出了：提高统计数据质量；增强中央银行主动搜集和分析信息的能力；完善我国的宏观经济模型；疏通货币政策传导渠道等相关政策建议。
[Abstract]:As an advance indicator to judge whether the economy is stable, the inflation rate directly or indirectly affects the macroeconomic indicators such as wages, interest rates, exchange rates and so on. Since the 1990s, keeping the inflation rate at a reasonable level to promote the smooth development of the economy has gradually become the central goal of the monetary policy of the central banks of various countries. In this context, inflation forecasting, as a new intermediate target of monetary policy, plays an increasingly important role in improving the effectiveness of monetary policy and stabilizing prices, especially in countries with inflation targeting system. At present, although there are many methods of predicting inflation, there are some shortcomings, and the decision of the central bank is not only based on the results of a model, therefore, a new forecasting method is constructed. It is of theoretical and practical significance to provide a comparative reference for central bank decision-making. This article follows from the theory elaboration, the demonstration analysis to the countermeasure suggestion train of thought unfolds. As a new forecasting method, neural network has been widely used in foreign countries because of its nonlinearity, data-driven and high stability. Although there are many researches in China, it is less in the field of inflation prediction. Therefore, based on the research results at home and abroad, this paper studies the applicability of neural network in China's inflation prediction from the perspective of theory and practice by combining qualitative analysis with quantitative analysis. In this paper, the basic theory of neural network and inflation prediction is introduced briefly, and the related concepts are explained. Then it lists the workflow and characteristics of some countries with rich experience in inflation forecasting, and compares and analyzes the shortcomings of our country in this respect. Based on the monthly data from March 2005 to December 2011, a BP neural network model is constructed to predict the inflation in China. The empirical results show that the optimized network can forecast CPI for at least 6 months according to the existing data, and the unoptimized network also gives a more reasonable forecast result. Finally, the paper puts forward how to improve the accuracy of the inflation forecast of the central bank: to improve the quality of statistical data, to strengthen the ability of the central bank to collect and analyze information actively, to perfect the macroeconomic model of our country; Dredging monetary policy transmission channels and other relevant policy recommendations.
【学位授予单位】：湖南大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：F822.5;TP18

【参考文献】