组合预测中单项预测模型的选择研究
发布时间：2018-10-29 09:06
【摘要】：随着社会经济的发展,统计预测技术也在逐渐成熟当中。目前,越来越多的现象被以数据这种具体的形式体现出来。因此,统计预测技术得以被应用于多个领域。从农业到工业,从人口到经济,从能源到社会发展,从交通运输到医疗卫生,统计预测技术几乎无处不在。针对同一个社会经济预测问题,由于决策者考虑的问题不同,通常会选择不同的预测方法。但是在处理数据的过程中单纯的单项预测模型在某些情况下已无法满足我们对预测精度的需求,组合预测开始越来越广泛的出现在实际应用当中。然而组合预测中单项预测模型的选择还存在需要继续研究的问题。本文即围绕如何选择并确定组合预测中的单项预测模型展开研究与讨论。本文首先阐明了论文的研究背景和研究意义,回顾国内外研究文献,提出本文研究的主要内容、研究框架和研究方法。其次,对于统计预测技术给出了进一步的论述,明确了预测工作的详细步骤,概括说明了单项预测模型和组合预测模型的相关理论,并对评价预测效果的指标体系做了介绍。然后,分别从定性和定量两个角度分析组合预测中单项预测模型的选择。定性角度的分析包括单项预测模型选择应遵循的原则、以及模型选择,即通过预测目的、历史数据变化的特征、预测模型的适用性初步确定单项预测模型的选择范围。定量角度的分析包括单项预测模型个数的确定以及权重的确定,从而得到适宜的组合预测模型。同时选取1994年到2013年中国私人汽车拥有量的历史数据,进行建模分析,验证单项预测模型选择步骤的合理性和可行性。最后从预测值和误差评价体系两个方面分析模型的预测结果,并为今后采用组合预测模型建模提出建议。
[Abstract]:With the development of social economy, statistical forecasting technology is gradually maturing. At present, more and more phenomena are reflected in the concrete form of data. Therefore, statistical prediction technology can be applied in many fields. From agriculture to industry, from population to economy, from energy to social development, from transportation to health care, statistical forecasting technology is almost everywhere. For the same socio-economic forecasting problem, different prediction methods are usually chosen because of the different problems considered by decision makers. However, in the process of processing the data, the single prediction model can not meet our demand for prediction accuracy in some cases, and the combination forecasting is becoming more and more widely used in practical applications. However, the choice of single prediction model in combinatorial forecasting still needs to be studied. This paper focuses on how to select and determine the single prediction model in combination forecasting. In this paper, the background and significance of the research are first expounded, and the domestic and foreign research literature is reviewed, and the main contents, research framework and research methods of this paper are proposed. Secondly, the paper gives a further discussion on statistical forecasting technology, clarifies the detailed steps of forecasting work, summarizes the relevant theories of single prediction model and combination forecasting model, and introduces the index system for evaluating the effect of prediction. Then, the selection of single prediction model in combination prediction is analyzed from qualitative and quantitative perspectives. The qualitative analysis includes the principles to be followed in the selection of the single prediction model and the selection of the model, that is, the selection range of the single prediction model is preliminarily determined by the prediction purpose, the characteristics of historical data change and the applicability of the prediction model. The quantitative angle analysis includes the determination of the number of single prediction models and the determination of weights, so as to obtain a suitable combination prediction model. At the same time, the historical data of Chinese private car ownership from 1994 to 2013 are selected to model and analyze, to verify the rationality and feasibility of the single prediction model selection step. Finally, the prediction results of the model are analyzed from two aspects: the prediction value and the error evaluation system, and some suggestions are put forward for the modeling of the combined forecasting model in the future.
【学位授予单位】：长安大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：F426.471
本文编号：2297254
[Abstract]:With the development of social economy, statistical forecasting technology is gradually maturing. At present, more and more phenomena are reflected in the concrete form of data. Therefore, statistical prediction technology can be applied in many fields. From agriculture to industry, from population to economy, from energy to social development, from transportation to health care, statistical forecasting technology is almost everywhere. For the same socio-economic forecasting problem, different prediction methods are usually chosen because of the different problems considered by decision makers. However, in the process of processing the data, the single prediction model can not meet our demand for prediction accuracy in some cases, and the combination forecasting is becoming more and more widely used in practical applications. However, the choice of single prediction model in combinatorial forecasting still needs to be studied. This paper focuses on how to select and determine the single prediction model in combination forecasting. In this paper, the background and significance of the research are first expounded, and the domestic and foreign research literature is reviewed, and the main contents, research framework and research methods of this paper are proposed. Secondly, the paper gives a further discussion on statistical forecasting technology, clarifies the detailed steps of forecasting work, summarizes the relevant theories of single prediction model and combination forecasting model, and introduces the index system for evaluating the effect of prediction. Then, the selection of single prediction model in combination prediction is analyzed from qualitative and quantitative perspectives. The qualitative analysis includes the principles to be followed in the selection of the single prediction model and the selection of the model, that is, the selection range of the single prediction model is preliminarily determined by the prediction purpose, the characteristics of historical data change and the applicability of the prediction model. The quantitative angle analysis includes the determination of the number of single prediction models and the determination of weights, so as to obtain a suitable combination prediction model. At the same time, the historical data of Chinese private car ownership from 1994 to 2013 are selected to model and analyze, to verify the rationality and feasibility of the single prediction model selection step. Finally, the prediction results of the model are analyzed from two aspects: the prediction value and the error evaluation system, and some suggestions are put forward for the modeling of the combined forecasting model in the future.
【学位授予单位】：长安大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：F426.471
【相似文献】
相关硕士学位论文 前1条
1 于秀伟;组合预测中单项预测模型的选择研究[D];长安大学;2015年
，本文编号：2297254
本文链接：https://www.wllwen.com/qiyeguanlilunwen/2297254.html
