【作者】 程玲华；

【导师】 陈华友；

【作者基本信息】 安徽大学 ， 应用数学， 2006， 硕士

【摘要】 由于社会经济系统的复杂性，采用单个预测模型进行预测存在预测的风险，于是提出了组合预测方法。二十世纪九十年代以来，组合预测方法成为国内外预测界研究的热点课题，组合预测方法己取得一系列的研究成果。但是组合预测方法产生的历史并不长，其模型和理论研究还不完善，有必要进一步加强研究。本文在现有文献的基础上，提出了新的组合预测模型，研究它们的一些性质。这对进一步丰富组合预测方法，发展预测学科的理论以及加强科学管理等方面具有重要的意义。 第一章对定性预测方法和定量预测方法进行了评价，综述了组合预测方法研究的国内外现状。 第二章介绍了基于误差指标的组合预测模型。包括常用的非最优正权组合预测模型，加权算术平均最优组合预测模型研究和几种基于信息集结算子的组合预测模型。介绍了组合预测的权系数计算方法。 第三章作者重点针对基于Theil不等系数的加权调和平均组合预测模型和基于Theil不等系数的加权几何平均组合预测模型，分别提出了新的优性组合预测、预测方法优超和冗余度等概念；指出了简单平均组合预测方法至少是非劣性组合预测；研究其非劣性组合预测和优性组合预测存在的条件、组合预测方法冗余信息的存在性及其判定。 第四章针对基于Theil不等系数的加权调和平均组合预测模型和基于Theil不等系数的加权几何平均组合预测模型进行实例分析，计算结果令人满意。表明本文所提模型是有效的。更多还原

【Abstract】 Due to the complexity of social economic system, it is sure to encounter forecasting risks by employing traditional individual forecasting models. As a result, combination forecasting methods were proposed, which has become a topic of general interest in the fields of forecasting home and abroad since 1990s. A series of researches have been done on this subject and many papers have also been published. However, the models and theories of combination forecasting are incomplete due to its short history. Based on the literatures available, the author of this thesis proposes a new combination forecasting model, and discusses its properties, which contributes to enrich methods of combination forecasting, develop theories of forecasting science and strengthen scientific management.In chapter one, evaluations of qualitative and quantitative forecasting methods are given, and the domestic and overseas studies on combination forecasting are also summarized.In chapter two, the combination forecasting models based on the error index are introduced, which includes the non-optimal combination forecasting models, weighted arithmetic average optimal combination forecasting models and several combination forecasting models given on the basis of information set. Besides, the author offers the weight coefficient methods of calculation on combination forecasting in this part.Chapter three witnesses the proposal of some new concepts, namely, superior combination forecasting, dominant forecasting method and redundant measure, which are grounded on the systematic analysis on weighted harmonic average optimal combination forecasting models and weighted geometric average optimal combination forecasting models taking Theil coefficient into consideration. In addtion, the author argues that simple average combination forecasting methods are at least non-inferior and hence studies the sufficient conditions of existence about non-inferior and superior combination forecasting and the determining of redundant更多还原