湖北省工业企业创新投入产出的多元非线性回归模型研究
发布时间：2019-06-06 04:47
【摘要】：创新能力是科技创新驱动发展战略的根本,是国家竞争力的具体体现,决定了国民经济的长期动力,预示着未来几年的经济发展趋势。因此,对创新能力的研究具有十分重要的意义。对于政府职能部门和企业这两个可调控的主体,深入改革促进创新的前提是能正确地评估企业的创新能力。在此背景下,2012年我们参与并完成了《湖北省工业企业创新能力综合评价》项目,项目成果收录在《湖北省第二次RD资源清查论文集》中。 投入产出模型是目前经济、管理学科研究中的重要工具,被广泛应用于各类综合评价及预测中。实际的投入产出模型中,指标体系往往比较庞大复杂,模型的建立大多基于多元线性回归方法。但是多元线性模型的建立经常会遇到两类的问题,第一类是模型的选择,另一类是数据的可靠性。比如：异方差问题,我们往往需要先对变量进行转换,这样与之所匹配的模型就不再是多元线性模型；又或者数据存在异常值,多元线性模型会受异常值影响而产生较大偏差。在本文中,我们采取了多元非线性回归及加权多元非线性回归的方法解决上述问题,并得到了较为合理的结果,得到了相关专家的认可。 本文选取湖北省工业企业作为研究对象,数据来源于2010-2012三年的《湖北省科技统计年鉴》,在构建科学合理的创新能力评价指标体系的基础上,对湖北省工业企业的创新能力建立投入产出模型进行研究。本文具体内容安排如下： 第一章为绪论。论述了本文相关的研究背景、研究意义和研究现状,叙述了本项目的主要研究内容和方法以及研究重点与特色。 第二章为湖北省工业企业创新能力综合评价基本情况介绍,包括创新能力综合评价指标体系的构建和基本指标的说明。本文建立的创新能力投入产出模型是综合评价的重要组成部分,其投入产出指标选取具备一定的科学性和合理性。 第三章为多元非线性回归模型理论及其应用。介绍了非线性回归模型的一般理论方法。特别提出了加权多元非线性回归方法,并举例说明了该方法的可行性。 第四章为湖北省工业企业创新投入产出多元非线性回归模型的实现。首先利用一般多元线性回归的方法建立投入产出模型,针对某些指标建立多元线性回归模型的不合理情况,我们进行了相应的非线性化处理；其次根据残差图发现出现的异常值,采取人工剔除异常值,继而使用“去异常值”的数据构建投入产出模型；最后提出并实现了“利用残差距离定义加权权重”的加权多元非线性回归模型,一定程度上避免了“去异常值”过程中的主观性和复杂性,结果也比较理想。 第五章为结论与展望。综述了全文结论,并讨论了本文的加权非线性回归模型。实证分析加权非线性回归模型具有一定可行性和合理性,但是其相关的理论性质以及更优的权重设置方法都值得深入研究。
[Abstract]:Innovation ability is the foundation of the strategy driven by scientific and technological innovation, and the concrete embodiment of national competitiveness, which determines the long-term driving force of the national economy and indicates the trend of economic development in the next few years. Therefore, the study of innovation ability is of great significance. For the government functional departments and enterprises, the premise of in-depth reform to promote innovation is to correctly evaluate the innovation ability of enterprises. In this context, in 2012, we participated in and completed the "Comprehensive Evaluation of Innovation ability of Industrial Enterprises in Hubei Province". The results of the project are included in the second RD Resource inventory of Hubei Province. Input-output model is an important tool in the research of economy and management, which is widely used in all kinds of comprehensive evaluation and prediction. In the actual input-output model, the index system is often large and complex, and most of the models are based on multiple linear regression methods. However, the establishment of multivariate linear models often encounters two kinds of problems, the first is the choice of models, the other is the reliability of data. For example, in the heteroscedasticity problem, we often need to transform the variables first, so that the matching model is no longer a multivariate linear model, or if there are outliers in the data, the multivariate linear model will be affected by the outliers and produce a large deviation. In this paper, we adopt the methods of multivariate nonlinear regression and weighted multivariate nonlinear regression to solve the above problems, and obtain more reasonable results, which have been recognized by relevant experts. In this paper, Hubei industrial enterprises are selected as the research object, and the data come from the Hubei Science and Technology Statistical Yearbook from 2010 to 20123. on the basis of constructing a scientific and reasonable evaluation index system of innovation ability, The input-output model of innovation ability of industrial enterprises in Hubei Province is studied. The specific content of this paper is arranged as follows: the first chapter is the introduction. This paper discusses the related research background, research significance and research status, and describes the main research contents and methods, as well as the research focus and characteristics of this project. The second chapter introduces the comprehensive evaluation of innovation ability of industrial enterprises in Hubei Province, including the construction of comprehensive evaluation index system of innovation ability and the explanation of basic indicators. The input-output model of innovation ability established in this paper is an important part of comprehensive evaluation, and its input-output index selection is scientific and reasonable. The third chapter is the theory of multivariate nonlinear regression model and its application. The general theory and method of nonlinear regression model are introduced. In particular, a weighted multivariate nonlinear regression method is proposed, and an example is given to illustrate the feasibility of the method. The fourth chapter is the realization of multiple nonlinear regression model of innovation input-output in Hubei Province. Firstly, the input-output model is established by using the general multivariate linear regression method, and the corresponding nonlinear treatment is carried out in view of the unreasonable situation of establishing the multivariate linear regression model for some indexes. Secondly, according to the abnormal values found by residual map, the abnormal values are manually eliminated, and then the input-output model is constructed by using the data of "removing outliers". Finally, a weighted multivariate nonlinear regression model of "using residual distance to define weighted weight" is proposed and implemented, which avoids the subjectivity and complexity in the process of "de-abnormal value" to a certain extent, and the results are also satisfactory. The fifth chapter is the conclusion and prospect. The conclusion of this paper is reviewed, and the weighted nonlinear regression model is discussed. Empirical analysis of weighted nonlinear regression model is feasible and reasonable, but its related theoretical properties and better weight setting methods are worthy of further study.
【学位授予单位】：华中师范大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：F425;F273.1;O212.1
[Abstract]:Innovation ability is the foundation of the strategy driven by scientific and technological innovation, and the concrete embodiment of national competitiveness, which determines the long-term driving force of the national economy and indicates the trend of economic development in the next few years. Therefore, the study of innovation ability is of great significance. For the government functional departments and enterprises, the premise of in-depth reform to promote innovation is to correctly evaluate the innovation ability of enterprises. In this context, in 2012, we participated in and completed the "Comprehensive Evaluation of Innovation ability of Industrial Enterprises in Hubei Province". The results of the project are included in the second RD Resource inventory of Hubei Province. Input-output model is an important tool in the research of economy and management, which is widely used in all kinds of comprehensive evaluation and prediction. In the actual input-output model, the index system is often large and complex, and most of the models are based on multiple linear regression methods. However, the establishment of multivariate linear models often encounters two kinds of problems, the first is the choice of models, the other is the reliability of data. For example, in the heteroscedasticity problem, we often need to transform the variables first, so that the matching model is no longer a multivariate linear model, or if there are outliers in the data, the multivariate linear model will be affected by the outliers and produce a large deviation. In this paper, we adopt the methods of multivariate nonlinear regression and weighted multivariate nonlinear regression to solve the above problems, and obtain more reasonable results, which have been recognized by relevant experts. In this paper, Hubei industrial enterprises are selected as the research object, and the data come from the Hubei Science and Technology Statistical Yearbook from 2010 to 20123. on the basis of constructing a scientific and reasonable evaluation index system of innovation ability, The input-output model of innovation ability of industrial enterprises in Hubei Province is studied. The specific content of this paper is arranged as follows: the first chapter is the introduction. This paper discusses the related research background, research significance and research status, and describes the main research contents and methods, as well as the research focus and characteristics of this project. The second chapter introduces the comprehensive evaluation of innovation ability of industrial enterprises in Hubei Province, including the construction of comprehensive evaluation index system of innovation ability and the explanation of basic indicators. The input-output model of innovation ability established in this paper is an important part of comprehensive evaluation, and its input-output index selection is scientific and reasonable. The third chapter is the theory of multivariate nonlinear regression model and its application. The general theory and method of nonlinear regression model are introduced. In particular, a weighted multivariate nonlinear regression method is proposed, and an example is given to illustrate the feasibility of the method. The fourth chapter is the realization of multiple nonlinear regression model of innovation input-output in Hubei Province. Firstly, the input-output model is established by using the general multivariate linear regression method, and the corresponding nonlinear treatment is carried out in view of the unreasonable situation of establishing the multivariate linear regression model for some indexes. Secondly, according to the abnormal values found by residual map, the abnormal values are manually eliminated, and then the input-output model is constructed by using the data of "removing outliers". Finally, a weighted multivariate nonlinear regression model of "using residual distance to define weighted weight" is proposed and implemented, which avoids the subjectivity and complexity in the process of "de-abnormal value" to a certain extent, and the results are also satisfactory. The fifth chapter is the conclusion and prospect. The conclusion of this paper is reviewed, and the weighted nonlinear regression model is discussed. Empirical analysis of weighted nonlinear regression model is feasible and reasonable, but its related theoretical properties and better weight setting methods are worthy of further study.
【学位授予单位】：华中师范大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：F425;F273.1;O212.1
【相似文献】
相关期刊论文 前10条
1 马阳明;韦博成;;带约束非线性回归模型的影响分析[J];东南大学学报;1993年03期
2 吴传绪;非线性回归模型的应用[J];陕西工学院学报;1997年01期
3 郝海燕;杨瑞臣;;三角进化算法在非线性回归模型建立中的应用[J];承德石油高等专科学校学报;2009年01期
4 王蕊;董祥e，
本文编号：2494082
本文链接：https://www.wllwen.com/jingjilunwen/gongyejingjilunwen/2494082.html
