【作者】 曹灿；

【导师】 赵联文；

【作者基本信息】 西南交通大学 ， 概率论与数理统计， 2011， 硕士

【摘要】 失业问题始终是当今世界各国社会经济发展的重大问题,它既是综合性的经济问题,又是复杂的社会问题。同时,失业是宏观经济中特别重要的三个指标之一,因此研究我国城镇失业率具有积极的现实意义论文中的数据来自于中华人民共和国统计局、国家统计数据库、国研网与和讯网,分别应用多元回归分析和支持向量回归法对我国城镇失业率进行预测,最后得出基于支持向量回归法优于多元回归分析方法。文章的主要内容是：本文绪论介绍了论文的背景和研究意义,对失业率与支持向量机进行了系统研究。第二章介绍了相关的预测方法,分别从定性预测和定量预测进行了研究。在定量预测中主要介绍了时间序列预测、多元回归预测、灰色预测、人工神经网络预测、支持向量回归和组合预测。第三章介绍了支持向量机的核心理论。其内容主要包括机器学习、统计学习理论和支持向量机三个方面。支持向量机方法是基于统计学习理论的一种算法,其基本理论就是VC维和结构风险最小化原理。第四章应用多元线性回归法、非线性回归法、神经网络法、线性核函数的支持向量机法、高斯径向基核函数的支持向量机法对我国城镇失业率进行了实证分析。这五种方法从曲线拟合图、拟合相对误差、拟合精度三个方面进行比较,得出支持向量机回归法是一种较理想的曲线拟合法。支持向量机回归法选取线性核函数和高斯径向基核函数对我国城镇失业率进行拟合,其拟合精度分别是0.177%和0.195%。最后,文章应用上述五种方法对2010年至2015年我国城镇失业率进行预测,并得出其预测结果。更多还原

【Abstract】 Unemployment has always been a major social and economic development issue throughout the world. It is both a comprehensive economic problem, but also a complex social problem. Meanwhile, unemployment is one of three important indicators in microeconomic. Therefore, the study of urban unemployment rate has a positive and practical significance.Daters of the thesis which come from Bureau of Statistics People’s Republic of China, National statistical Database, Development Research Centre of the State Council Information Networks and Hexun Networks predict China’s urban unemployment rate using regression analysis and support vector regression methods.ChapterⅠintroduces the background and significance of the paper. It also does some research in unemployment rate and Support Vector Machines. ChapterⅡpresents the relevant forecasting methods, which are studied from the qualitative and quantitative prediction. The quantitative prediction contents Time Series prediction, Multiple Regression forecast, Grey Prediction, Artificial Neural Networks prediction, Support vector Regression and compound prediction. ChapterⅢdescribes the core of the support vector machines theory in three aspects, which shows from machine learning, statistical learning theory and support vector machines. Support vector machines method is based on an algorithm of statistical learning theory. The basic theory is VC dimension (Vapnik and Chervonenkis dimension) and structural risk minimization principle. ChapterⅣhas made an empirical analysis in China’s urban unemployment in the way of multiple linear regression, nonlinear regression, neural networks, support vector machine of linear kernel and Gaussian radial basis function. These five methods compares from three aspects,including the curve fitting, relative error and the fitting precision..Therefore,we come to a conclusion that the SVM regression method is an ideal curve fitting. Support vector regression selects linear kernel function and Gaussian Radial Basis Function kernel for its fitting. The fitting accuracy is of 0.177% and 0.195%. Finally, this paper forecasts China’s urban unemployment rate from 2010 to 2015 using those above five methods and obtained its results.更多还原