人工神经网络算法在GDP和CPI中的预测应用

发布时间：2018-02-20 16:43

本文关键词： GDP CPI BP神经网络算法主成分分析粒子群算法灰狼优化算法支持向量机　出处：《中北大学》2017年硕士论文　论文类型：学位论文

【摘要】：国际经济竞争越来越激烈,为保证我国在国际竞争中立于不败之地,必须保证宏观经济稳定发展。GDP增长率反映了国家经济状况的整体水平,CPI指数直接影响国民购买力。在制定宏观经济政策时,必须研究历史数据,从历史数据中寻求它们之间的内在关系,进而为制定合适的方针政策提供指导。GDP增长率与CPI指数的历史数据具有复杂的时间序列性和非线性性。人工神经网络算法具有良好的非线性拟合能力,在处理非线性问题时得到了广泛应用。BP神经网络算法和SVM算法作为人工神经网络中两种广泛使用的算法,都具有良好的非线性拟合能力,但也存在不足。针对不足之处,本文做了以下几方面的工作。1、针对BP神经网络算法容易陷入局部极小值,提出了PSO-BP模型算法。该方法主要利用PSO算法良好的全局寻优能力对BP算法的权值和阈值进行优化,从而避免BP算法陷入局部极小值。2、SVM算法中参数的选择直接影响模型的性能,所以如何选择恰当的参数至关重要。文中利用GWO灰狼算法良好的全局寻优能力对SVM的参数进行寻优,从而提高SVM模型的预测精度。3、PCA主成分可以实现降维,并保留原始数据的绝大部分信息,文中用PCA算法对实验数据进行降维,来提高模型的预测精度。通过PCA-PSO-BP模型对GDP增速和CPI指数进行拟合预测,实验结果表明PCA-PSO-BP模型比PCA-BP模型和PSO-BP模型具有更高的拟合精度和更小的均方误差。此外,文中也对PCA-GWO-SVM模型进行了GDP增长率和CPI指数的预测实验。结果表明PCA-GWO-SVM模型比PCA-SVM模型和GWO-SVM模型具有更高的拟合精度和更小的均方误差。通过PCA-PSO-BP模型与PCA-GWO-SVM模型的实验结果对比发现,PCA-GWO-SVM模型有更小的均方误差。
[Abstract]:International economic competition is becoming fiercer and fiercer. In order to ensure that our country is in an invincible position in international competition, It is necessary to ensure that the steady development of the macro economy... GDP growth rate reflects the overall level of the national economic situation. CPI directly affects the purchasing power of the people. In formulating macroeconomic policies, we must study historical data. Looking for the interrelationship between them from historical data, The historical data of CPI index and GDP growth rate have complex time series and nonlinearity. The artificial neural network algorithm has good nonlinear fitting ability. In dealing with nonlinear problems, BP neural network algorithm and SVM algorithm are widely used as two widely used algorithms in artificial neural network. Both of them have good nonlinear fitting ability, but they also have some shortcomings. In this paper, the following work has been done. 1. Aiming at BP neural network algorithm is easy to fall into local minima, the PSO-BP model algorithm is proposed. This method mainly uses the good global optimization ability of PSO algorithm to optimize the weight and threshold of BP algorithm. In order to avoid the BP algorithm falling into the local minimum. 2SVM algorithm parameters selection directly affect the performance of the model, so how to select the appropriate parameters is very important. In this paper, the good global optimization ability of GWO gray wolf algorithm is used to optimize the parameters of SVM. In order to improve the prediction accuracy of SVM model. 3PCA principal component can achieve dimensionality reduction, and retain most of the information of the original data. In this paper, we use PCA algorithm to reduce the dimension of experimental data. The experimental results show that PCA-PSO-BP model has higher fitting accuracy and smaller mean square error than PCA-BP model and PSO-BP model. The experimental results of GDP growth rate and CPI exponent of PCA-GWO-SVM model show that PCA-GWO-SVM model has higher fitting precision and smaller mean square error than PCA-SVM model and GWO-SVM model. The experimental results of PCA-PSO-BP model and PCA-GWO-SVM model show that the PCA-GWO-SVM model has higher fitting accuracy and smaller mean square error than PCA-SVM model and GWO-SVM model. The results show that the PCA-GWO-SVM model has smaller mean square error.
【学位授予单位】：中北大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：F124;F726;TP183

【相似文献】