基于PLS-SVM的重庆市经济发展状况研究

发布时间：2018-04-01 13:24

本文选题：经济预测　切入点：指标体系　出处：《重庆大学》2014年硕士论文

【摘要】：经济的增长能够促进城市各方面的发展，城市发展起来了又可以带动经济的发展。因此，研究城市的经济发展状况便有了举足轻重的意义。随着生活质量的提高，经济的大发展，人们开始关注起经济的走势，越来越多的学者投入到经济状况及其未来发展的研究热潮中去。然而经济系统是一个复杂且多变的系统，要探索清楚它的本质势必具有一定的难度。用于经济预测的变量通常较多，过多的变量就容易产生多重共线性的问题，为了消除该影响，出现了提取成分的思想。偏最小二乘（PLS）集合了相关分析、成分提取、回归拟合等众多优势，成为了现今研究的热点。近年来，机器学习开始活跃起来，新兴的支持向量机（SVM）从最初的解决分类问题扩展到回归拟合上去，展现出了其独特的优势，受到众人的追捧。本文首先介绍了城市经济预测的研究背景及相关知识。然后，本文开始介绍PLS和SVM的相关理论，为后文的研究作好了准备。之后开始进行实证分析。本文根据1997-2012年的相关数据建立了PLS-SVM模型，并预测出2013年的GDP值，根据预测误差来看，拟合效果比较理想。接着，本文又建立了与本模型思想相似的PCA-SVM模型和直接的SVM模型，用于与本模型进行比较分析。最终，，本文得出了PLS-SVM模型的拟合效果优于另两种模型的结论。本文具备良好的实用性和适用性，利于扩大推广。
[Abstract]:Economic growth can promote the development of all aspects of the city, urban development can also lead to economic development.Therefore, the study of the economic development of the city has a pivotal significance.With the improvement of quality of life and the great development of economy, people begin to pay attention to the trend of economy, and more and more scholars are engaged in the upsurge of research on economic situation and its future development.However, the economic system is a complex and changeable system, it must be difficult to explore its essence.There are usually many variables used in economic prediction, and too many variables are easy to produce multiple collinear problems. In order to eliminate this effect, the idea of extracting components has emerged.Partial least squares (PLS), which has many advantages, such as correlation analysis, component extraction, regression fitting and so on, has become a hot research topic.In recent years, machine learning has become active, and the new support vector machine (SVM) has expanded from the original classification problem to regression fitting, showing its unique advantages, which has been sought after by many people.This paper first introduces the research background and related knowledge of urban economic forecasting.Then, this paper begins to introduce the related theories of PLS and SVM, and prepares for the later research.Then the empirical analysis began.In this paper, the PLS-SVM model is established based on the relevant data from 1997-2012, and the GDP value of 2013 is predicted. According to the prediction error, the fitting effect is satisfactory.Then, the PCA-SVM model and the direct SVM model, which are similar to this model, are established, which are used to compare and analyze the model.Finally, this paper draws the conclusion that the fitting effect of PLS-SVM model is better than that of other two models.This article has the good practicability and the applicability, is advantageous to expand popularizes.
【学位授予单位】：重庆大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：F224;F127

【参考文献】