基于关联度的优性区间型组合预测模型研究

发布时间：2018-01-08 09:14

本文关键词：基于关联度的优性区间型组合预测模型研究　出处：《安徽财经大学》2017年硕士论文　论文类型：学位论文

【摘要】：由于数据样本具有异质性,因此针对研究对象的预测方法应进行差异化处理。有时尽管找到的预测方法整体上已较为匹配,但是在某些特殊时间点的预测误差仍然较大。组合预测则是通过组合几种不同的单项预测方法,然后附上权重系数,进而建立组合预测模型,充分提取各个预测方法中的有效信息,吸取它们的优点,规避它们的缺点,提高预测精度。建立组合预测模型的关键问题就是如何构建适当的目标准则和信息集成算子,综合衡量各个预测方法,求出各单项方法在组合预测中的权重,进而得到更加准确的预测值。目前,各类组合预测方法大都运用于实数样本数据。然而,在实际的工作及生活中,人们对研究对象进行描述时会面临数据本身的不确定性、观测者主观思维的模糊性等问题,使得人们难以用较为准确的数据对研究对象进行描述。针对上述这些问题,文章尝试将实数序列换成区间数序列,使得组合预测方法更有意义。为了探究区间型数据的研究对象,全文构建了六个区间数组合预测模型,并且通过理论证明和实例验证对其有效性进行了说明。本文的主要内容如下:(1)分别对区间数的中点和半径运用灰色关联度指标作为目标准则,构建基于灰色关联度的优性区间型组合预测多目标模型,偏好系数被引入模型中,从而实现多目标的单一化,使得模型得以计算,具体包括定权系数的组合预测模型和引入GIOWA算子的变权系数的组合预测模型,研究了定权系数模型的有效性理论:非劣性区间数组合预测的理论研究、组合预测中是否存在冗余单项方法的判断定理、冗余信息的判定定理,同时选取某省社会保障水平适度区间值作为样本,实例运算并验证了两类模型的有效性。(2)构建新的区间数关联度指标作为目标准则,分别建立基于区间数关联度的优性组合预测模型,基于区间数关联度及GIOWA算子变权重的组合预测模型,研究了定权重模型的有效性理论:非劣性区间数组合预测的理论研究、组合预测中是否存在冗余单项方法的判断定理、冗余信息的判定定理,同时选取某省社会保障水平适度区间值作为样本,实例计算验证了两类模型的有效性。(3)将区间数转变为二元联系数,构建新的二元联系数关联度指标作为目标准则,分别建立基于联系数关联度的优性组合预测模型,基于联系数关联度与GIOWA算子变权重的组合预测模型,研究了定权系数模型的有效性理论:非劣性区间数组合预测的理论研究、组合预测中是否存在冗余单项方法的判断定理、冗余信息的判定定理,同时使用某省社会保障水平适度区间值,实例运算并验证了两类模型的有效性。
[Abstract]:Because the data sample is heterogeneous, so the prediction method based on the research object should be differentiated. Sometimes although the overall prediction method found has been more, but in some special time point of the forecast error is still large. The combination forecast is forecast method by combining several different items, then attach weights, and then set up the combination forecasting model, fully extract effective information of each forecasting method, from their advantages and avoid their disadvantages, improve the prediction accuracy. The establishment of how to build the objective criteria and information integration operator appropriate key combination forecasting model, a comprehensive measure of each prediction method, calculate the weight of each method in the combination forecasting then, to get a more accurate prediction value. At present, all kinds of combination forecasting methods are applied to real sample data. However, in the actual Work and life, to describe the object of study people will face data uncertainty, fuzzy observer problems such as subjective thinking, making it difficult to describe the research object with accurate data. In view of the above questions, this article will try to change the real sequence of interval number sequence, the combination forecasting method is more in order to explore the research object. The interval data, this paper constructs six interval combination forecasting model, and its effectiveness is illustrated through theoretical proof and example verification. The main contents of this paper are as follows: (1) of interval number and the midpoint radius using gray correlation index as the objective criterion, construction optimal interval combination grey prediction model based on multi-objective, preference coefficient is introduced in the model, the single achieved the goal, so that the model can be The specific combination of calculation, combination forecasting model including fixed weight coefficient and introducing GIOWA operator variable weight coefficient prediction model, researches the theory of effective weighting coefficient model: theoretical study of non inferiority interval combination forecasting, judging whether there is redundant single theorem of combination forecasting method, theorem of redundant information. At the same time, choose a moderate social security level interval value as the sample, the calculation and verify the validity of the two models. (2) to construct interval correlation indicators as a new target, optimal combination forecasting models are established based on correlation degree of interval number, interval number combination relation degree and GIOWA operator variable weight prediction based on the model, study the theory of the validity of the weight model: a theoretical study of noninferior combination forecasting interval, the existence of redundant single method of combination forecast the judgment theorem, redundancy Theorem of additional information, and choose a moderate social security level interval value as the sample calculation to verify the effectiveness of the two model. (3) the interval number into two element connection number, set two element connection number new index as the objective criterion, optimal combination forecasting models are established connection number based on correlation, combination of connection number correlation and GIOWA operator variable weight prediction model based on the theory research, effective weighting coefficient model: theoretical study of non inferiority interval combination forecasting, judging whether there is redundant single theorem of combination forecasting method, theorem of redundant information, at the same time the use of a society the appropriate level of protection interval value calculation, and verify the validity of the two models.

【学位授予单位】：安徽财经大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：F224

【相似文献】