一种改进的RBF神经网络对县区级政府编制总量预测的研究

发布时间：2018-06-21 15:18

本文选题：RBF + 神经网络　；参考：《信阳师范学院》2015年硕士论文

【摘要】：我国现在正处于政府机构改革、政府职能转变的关键时期,机构和编制的决策必须适应社会主义市场经济的框架。而现阶段,仍有很多编制管理部门依然沿用趋势分析法、经验比例法等“老方法”进行编制预测,造成了编制核定工作的人为性、随意性、主观性,以及政府各部门之间编制申请增加、行政编制虚多、互相攀比等问题。因此,研究新形势下的编制总量预测方法就十分有意义。目前编制总量预测的主要方法有：经验比例法、趋势分析法、回归分析法、马尔可夫法、灰色预测方法等,国内已有专家学者对编制预测这个问题进行了探索性的研究,从政治、经济、管辖面积、管辖人口、文化传统、道德风尚等因素分析,建立了政府人员配置模型。本文提出了一种改进的RBF网络算法,并结合当地实际,选取河南省信阳市狮河区(政府级别为县区级)作为研究对象,将经济、民生等相关指标一并纳入到编制核定的指标体系中来,如：政府辖区的面积、人口、行政区划、国内生产总值、工农业产值、财政收人、行政经费等。选取狮河区历年的人口规模(人)X1,经济水平(万元)X2,地域面积(平方公里)X3,财政收入(亿元)X4,民间非政府组织X5,公职人员年龄占比分布X6,公职人员学历占比分布X7作为训练样本进行训练学习对泖河区政府编制总量进行了预测。本文的创新点主要体现在如下两个方面。 1.提出了基于RBF神经网络的县区级地方政府编制总量的预测方法。 2.对传统RBF网络模型在如下两方面进行了改进： (1)优化宽度σ取值。以往在确定函数中心宽度参数σ时仅根据经验进行学习,本文引入了GCV准则进一步优化宽度参数σ。 (2)对RBF网络进行子网络化优化处理。预测结果表明：基于改进的RBF网络算法对县(区)级地方政府编制总量的预测,比传统的编制总量预测方法误差更小,表现出精度更高的预测效果。这种编制总量预测方法比趋势分析法、经验比例法等“老方法”的系统性和科学性更强,因此,地方政府及编制管理部门在进行编制决策时能够以此作为重要依据,具有一定的推广应用价值。
[Abstract]:Our country is now in the government organization reform, the government function changes the key period, the organization and the establishment decision must adapt to the socialist market economy frame. However, at the present stage, there are still many establishment management departments that still use "old methods" such as trend analysis, experience-proportional methods, etc., to carry out compilation and prediction, resulting in the artificial, arbitrary and subjective nature of the compilation and approval work. As well as government departments between the establishment of applications, the administrative establishment of more false, compared with each other, and so on. Therefore, it is of great significance to study the method of forecasting the total amount of compilation under the new situation. At present, the main methods of compiling total forecast are: empirical proportion method, trend analysis method, regression analysis method, Markov method, grey forecast method and so on. This paper analyzes the factors of politics, economy, jurisdiction area, governing population, culture tradition, moral custom and so on, and establishes the model of government staffing. In this paper, an improved RBF network algorithm is proposed, and the Shihe District of Xinyang City, Henan Province (government level is county level) is selected as the research object. People's livelihood and other related indicators are incorporated into the approved index system, such as: the area of government jurisdiction, population, administrative divisions, GDP, industrial and agricultural output value, financial revenue, administrative funds, and so on. Select the population scale of Shihe District in the past years (people X 1, economic level (10 000 yuan x 2, geographical area x 3 square kilometers, revenue 1 million yuan X 4, non-governmental organizations X 5, age ratio of public officials X 6, educational background of public officials x 6%) X 7 was used as a training sample to predict the total amount of government establishment in Maohe district. The innovation of this paper is mainly reflected in the following two aspects. 1. Based on RBF neural network, the prediction method of the total amount of local government at county and district level is put forward. 2. The traditional RBF network model is improved in the following two aspects: 1) optimizing the width 蟽. In the past, the function center width parameter 蟽 was studied only according to experience. In this paper, the GCV criterion is introduced to further optimize the width parameter 蟽. The prediction results show that the prediction of the total amount of local government at the county (district) level based on the improved RBF neural network algorithm is less than that of the traditional method, and the prediction effect is higher than that of the traditional method. This method is more systematic and scientific than the old methods, such as trend analysis, experiential proportion, etc. Therefore, local governments and establishment management departments can take this as an important basis for making decisions. It has certain value of popularization and application.
【学位授予单位】：信阳师范学院
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：D630;TP18

【参考文献】