经济时间序列的趋势分析和实证研究

发布时间：2018-02-28 03:08

本文关键词： 经济时间序列 ARIMA模型支持向量机混合模型　出处：《首都经济贸易大学》2012年硕士论文　论文类型：学位论文

【摘要】：经济时间序列数据反映了经济各个方面的运行状况，，对其进行正确分析具有重要的意义。经济时间序列具有不稳定、复杂和难以预测的特征，分析方法包括传统的计量经济学方法，统计学方法和近几年发展起来的机器学习方法，这些方法都具有一定的优点，也存在一些不足，计量经济学方法对假设条件的要求比较严苛，神经网络等方法容易发生过拟合等问题。为了进一步提高趋势分析的准确性，本文将计量经济学中的ARIMA模型与支持向量机模型相结合，由于ARIMA模型在线性时间序列的分析中有较高的准确性，而支持向量机模型在非线性时间序列建模上具有较好的能力，所以将两者结合，形成混合模型，利用混合模型对经济时间序列进行分析。本文首先选取社会消费品零售总额的时间序列进行实证研究，社会消费品零售总额反映社会商品购买力的实现程度和零售市场的规模状况。在计量经济学建模时，根据其扰动项同方差的特点选择了ARIMA模型，再利用支持向量机模型对其进行研究，最后，引入ARIMA与支持向量机混合模型，将ARIMA模型得到的估测值与支持向量机模型对残差建模得到的估测值相加，得到混合模型结果。对三种模型的结果进行比较，验证混合模型的效果最佳。为了验证混合模型的准确性和应用范围，本文还选取了上证综合指数的时间序列进行分析，同样采取三种模型进行分析研究。最后得出结论：混合模型具有很好的拟合效果和估测精度。
[Abstract]:The data of economic time series reflect the operation of various aspects of economy, and it is of great significance to analyze it correctly. Economic time series are characterized by instability, complexity and unpredictability. The analytical methods include traditional econometric methods, statistical methods and machine learning methods developed in recent years. The requirements of econometrics methods on the assumption conditions are strict, and the neural network methods are prone to the problems of over-fitting and so on. In order to further improve the accuracy of trend analysis, this paper combines the ARIMA model in econometrics with the support vector machine model, because the ARIMA model has a high accuracy in the analysis of linear time series. The support vector machine (SVM) model has a good capability in nonlinear time series modeling, so it combines the two models to form a hybrid model and uses the hybrid model to analyze the economic time series. This paper first selects the time series of total retail sales of consumer goods to conduct an empirical study. The total amount of retail sales of consumer goods reflects the degree of realization of purchasing power of social goods and the scale of retail market. According to the characteristic of the same square difference of the disturbance term, the ARIMA model is selected, and then the support vector machine model is used to study it. Finally, the mixed model of ARIMA and support vector machine is introduced. The results of mixed model are obtained by adding the estimated values of ARIMA model and support vector machine model to the residual model, and the results of the three models are compared to verify the best effect of the hybrid model. In order to verify the accuracy and application range of the hybrid model, this paper also selects the time series of Shanghai Composite Index to analyze. Finally, the conclusion is drawn that the mixed model has good fitting effect and estimation accuracy.
【学位授予单位】：首都经济贸易大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：F224;F832.51;F726

【参考文献】