基于特征向量提取的核回归组合预测模型及在我国煤炭消费预测中的应用研究
详细信息    本馆镜像全文|  推荐本文 |  |   获取CNKI官网全文
摘要
煤炭是我国能源体系中最重要的物质基础之一,煤炭消费预测对我国的宏观经济健康发展有着重要作用。科学的煤炭消费预测有助于合理、有效地安排国民生产。随着我国经济的发展和工业化进程的深入,煤炭的需求量不断上升,供需之间矛盾日益加剧,因此,利用科学合理的研究方法,对今后数年我国在煤炭领域的消费数量进行预测会推动经济的不断进步。
     在煤炭消费预测领域中,存在许多的单项预测模型,同时学者们也在对单项预测方法进行不断的改进,从而可以地更好刻画、模拟和预测。目前,煤炭消费预测通常采用灰色预测法,回归分析法、趋势外推预测法以及ARMA预测模型等,但是这些方法由于自身的缺陷在预测中存在一定的不足。所以任何单一模型在都有它自身的优缺点,完美无缺的预测模型是不存在的,而预测对象错综复杂,数据变化的趋势表现出较强的随机波动性,所以在现实的预测研究中,预测的结果会与实际数据有比较大的偏差。
     组合预测方法克服了单一预测方法的不足与局限性,因为单一预测方法往往侧重的是一个方面而忽略了其他有用的信息,组合预测方法是将不同的单一预测模型集结在一个特定环境中,通过一定的技术手段赋予各个单一预测模型不同的权重系数,这样可以全面、科学、有效地进行预测,达到提高预测精度的目的。
     目前组合预测方法的类型繁多、枚不胜举,其中最优组合预测是当前研究的一个重要内容,但是最优组合预测存在着权重系数可能为负的局限性,而非负权重组合预测的虽然避免了权重为负的可能性,但是实证效果却不及最优组合预测,为了缓解这二者的矛盾,于是文章提出了基于特征向量提取的核回归组合预测方法,即努力在最优组合预测的基础上来寻找权重系数都为正的情况。实证部分将1958年至2010年我国煤炭的历史消费数据运用到煤炭消费的实证研究中,通过与其他方法预测进行比较得出该组合预测方法存在一定的优势。
     为了说明基于核回归组合预测模型更适用于我国煤炭的消费预测,文章先分析了组合预测研究领域中的研究背景及意义,接着详细介绍常用煤炭消费的单一预测模型性质与局限性,最后提出基于特征向量提取的核回归组合预测模型,并在实证研究中得出基于核回归组合预测模型的预测结果优于其他预测模型,故在煤炭消费预测中核回归组合预测模型具有更强的适应性。
Coal energy system in China is one of the most important material basis, coalconsumption prediction on China's macro-economic health has an important role.Scientific prediction of coal consumption contribute to rational and effectivearrangements for national production. With the deepening of China's economicdevelopment and industrialization, increasing demand for coal, growing contradictionbetween supply and demand and, therefore, take advantage of scientific researchmethods, in China in the next few years in the field of coal consumption prediction willpromote economic progress.
     In field of coal consumption prediction, prediction model of many items, whilescholars are for single prediction method for continuous improvement, so that you canbetter characterization, simulation and forecasting. At present, the grey forecastingmethod of coal consumption prediction is usually used, method of regression analysis,trend prediction method and ARMA predictive model, but these methods due to theirown shortcomings in certain deficiencies in the forecast. So no single model has its ownadvantages and disadvantages, there is no perfect prediction model, and predict complex,changes in data trends showed strong random fluctuations, so in reality the prediction,prediction of deviation of the results will be compared with the actual data.
     The combination forecast method overcome has single forecast method ofinsufficient and limitations, because single forecast method often focused on of is a areaand ignored has other useful of information, combination forecast method is willdifferent of single forecast model build-up in a specific environment in the, by must oftechnology means gives all single forecast model different of weight coefficient, so canfull, and science, and effective to for forecast, reached improve forecast precision ofpurpose.
     Currently combination forecast method of type range, and pieces be lift, whichoptimal combination forecast is current research of a important content, but optimal combination forecast exists with weight coefficient may for negative of limitations, andnon-negative weight combination forecast of although avoid has weight for negative ofpossibilities, but empirical effect is than optimal combination forecast, to mitigation thisboth of contradictions, so article made has based on features vector extraction of nuclearregression combination forecast method, Efforts in looking for the basis of optimalcombination forecasting of weighting coefficients are positive. Empirical sectionbetween1958and historical consumption data of coal to the coal consumption in China:an empirical study, through comparison with other methods to forecast that thecombined forecast methods present certain advantages.
     To description based on nuclear regression combination forecast model moreapplies Yu in China coal of consumption forecast, article first analysis has combinationforecast research area in the of research background and the significance, then moredescribes common coal consumption of single forecast model nature and limitations,last made based on features vector extraction of nuclear regression combination forecastmodel, and in empirical research in the came based on nuclear regression combinationforecast model of forecast results better than other forecast model, so in coalconsumption forecast in the nuclear regression combination forecast model has morestrong of adaptability.
引文
[1]薄志毅.滚动组合预测模型研究及其应用.[J]测绘速报.2009.
    [2]汪同三、张涛.组合预测:理论、方法及应用.[M].第一版.北京:社会科学文献出版社,2008,38-41。
    [3] Bates J M,Granger C W J. The combination of forecasts.[J].Operational ResearchQuarterly,196920:451~468
    [4]陈华友.组合预测方法有效性理论及其应用.[M].第一版.北京:科学出版社.2008,40-42.
    [5] Newbold,P.,and Granger,C.W.J.“Experience with forecasting univariate time series and thecombination of forecasting”.[J].Journal of the Royal Statistical Society.1974.Vol.137,No.2
    [6] Makridakis,S.et al,(1983):The accuracy of major ectrapolation methods.Vol,111-113
    [7] G. Peter Zhang, Time series forecasting using a hybrid ARIMA and neural network model,[J].Neurocomputing.2003,50:159~175
    [8] Huseyin Ince, Theodore B. Trafalis, A hybrid model for exchange rate prediction.[J] DecisionSupport Systems,2006,42:1054~1062
    [9] Zhongsheng Hua, Bin Zhang, A hybrid support vector machines and logistic regressionapproach for forecasting intermittent demand of spare parts.[J]Applied Mathematics AndComputation,2006,181:1035~1048
    [10] Granger,Amanathan r, Hybrid neural network models for hydrologic time seriesforecasting.[M] Applied Soft Computing,2006
    [11]唐小我.最优组合预测的计算方法.[J]管理现代化,1991(1)
    [12]文新辉、陈开周:一种新的最优组合预测方法及其应用[J].西安电子科技大学.1998
    [13]周传世:加权几何平均组合预测模型及其应用[J].数理统计与管理,1995(2)
    [14]杨广喜:经济决策与组合预测理论,武汉大学出版社,1998,46-48
    [15]唐小我.预测理论及其应用.[M]成都:电子科技大学出版社,1992
    [16]王明涛:预测方法的有效性分析[J].预测.1994(6)
    [17]唐小我、曹长修、金德运:组合预测最优加权系数向量的进一步研究[J].1994(2)
    [18]周传世:非线性组合预测模型及其最优权的确定[J].预测,1994(2)
    [19]曾勇、唐小我:线性规划在非负权重最优组合预测计算中的应用[J].预测,1994(3)
    [20]王应明.调和平均组合预测方法的进一步研究[J].系统工程与电子技术,1998,20(9):42-45.
    [21]王应明.基于相关性的组合预测方法研究[J].预测,2002,21(2):58-62.
    [22]王应明.广义加权几何平均组合预测方法研究[J].厦门大学学报(自然科技版),1998,37(1):6-10.
    [23]何芳,陈收.基于扩展Kalman滤波的神经网络学习算法在股票预测中的应用[J].中国管理科学,2003(06):75-79.
    [24]陈华友,盛昭瀚,刘春林.调和平均的组合预测方法之性质研究[J]。系统工程学报,2004,19(6):620-624.
    [25]陈华友.基于相关系数的优性组合预测模型研究[J]。系统工程学报,2006,21(4):352-360.
    [26]陈华友.广义加权算术平均组合预测法的最优化理论基础及性质[J].系统工程理论与实践,2003,23(4):37-41.
    [27]陈华友、侯定丕:基于预测有效度的优性组合预测模型研究[J].中国科技大学学报,2002(2)
    [28]曾勇、唐小我、郑维敏:基于斯坦规则和误差校正的组合预测模型[J].管理工程学报,2001(2)
    [29]张伟.基于状态转移的人民币汇率时变组合预测研究[D].重庆师范大学,2010
    [30]王黎明.金融时间序列理论及其应用研究.[D].上海财经大学,2004
    [31] G. Peter Zhang, Time series forecasting using a hybrid ARIMA and neural network model,[J].Neurocomputing.2003,50:159~175
    [32] Ruey S.Tasy著,潘家柱译,金融时间序列分析,北京:机械工业出版社,2006
    [33]胡光宇.战略:预测与决策[M].北京:清华大学出版社,2005.
    [34]丛春霞,李秀芳.灰色预测在股票价格指数预测中的应用[J].中国统计,2000(5):15-17.
    [35]常学将陈敏王明生,时间序列分析[M].北京:高等教育出版社,1993:254-295
    [36]王居平.一种新的图书购置经费的组合预测方法.[J].情报杂志,2007
    [37]卫丽君.基于GM-BP模型的交通事故预测[D].西安科技大学,2011
    [38]陆剑锋.江苏省“三废”排放实证分析与研究.[D].南京航空航天大学,2007
    [39]穆昭光.灰色预测模型在江苏省GDP预测中的运用分析.[J].现代商贸工业,2009(22)
    [40]邓晓刚,田学民.基于特征向量提取的核回归建模方法研究.[J.]控制工程,2010(4)
    [41]任少龙.基于小波神经网络的装备研制消费预测与控制技术研究.[D].北京理工大学,2002

© 2004-2018 中国地质图书馆版权所有 京ICP备05064691号 京公网安备11010802017129号

地址:北京市海淀区学院路29号 邮编:100083

电话:办公室:(+86 10)66554848;文献借阅、咨询服务、科技查新:66554700