基于随机森林和规则集成法的酒类市场预测与发展战略
|
摘要
酒业是我国的传统产业，在国民经济发展中具有举足轻重的地位和作用。对于酒类企业，市场和利润是决定其生存和发展的根本。本文利用秩相关系数和随机森林法分析各种酒类产量及影响因素之间的相关性，并基于ARMA-GARCH模型和规则集成算法预测了天津津酒集团的利润，最后提出津酒集团未来发展战略。
根据全国各省市历年白酒、啤酒、葡萄酒的产量与各省市人口及城镇居民人均可支配收入数据，计算以上数据两两之间的Spearman和Kendall秩相关系数，结果表明啤酒、葡萄酒、人口和收入都影响着白酒产量。因此将白酒产量当作因变量，其它4个指标当作自变量，首次利用随机森林法拟合它们之间的函数关系，然后基于此拟合函数计算其它指标对于白酒产量的影响程度，并做出了偏相关图形。
通过津酒集团1986年1月到2006年12月20年每月的利润数据，首先根据数据的平稳性和ARCH效应检验结果，首次确定利润一阶差分均值采用线性ARMA模型，方差采用广义条件异方差GARCH模型。再通过检验残差的独立性，最后确定出ARMA(1,14)-GARCH(1,1)模型，极大似然估计参数，从而预测出津酒集团未来利润。
按照津酒集团利润数据和相同时段1986年1月到2006年12月每月产量、收入和税金。把利润当成因变量，产量、收入和税金当做自变量，首次利用规则集成法拟合它们之间的函数关系。采用GARCH模型预测自变量――产量、收入和税金未来对应数值，将预测出的自变量数值再代入拟合函数，从而预测未来利润。并用2007年1至6月的利润值进行检验，结果表明规则集成法预测误差更小。
通过全国和天津白酒行业现状、市场特征和发展趋势的分析，根据本文研究的啤酒、葡萄酒及白酒产量的影响程度和白酒利润预测结果，指出天津津酒集团面临机遇、挑战和发展瓶颈，提出津酒集团发展目标、思路和战略规划：从构建三大平台、提升六大能力、整合公司治理结构与集团化管理模式等方面的实施方案，从而为津酒集团的战略腾飞指明方向。
Liquor product is traditional industry in China. It plays an important role in na-tional economic development. Market and profit are the base for its surviving and devel-oping. The dependence between liquor output and their influential factors was analyzedby rank correlation coeffcient and Random Forest. The profit of Tianjin Jinjiu Groupwas predicted by ARMA-GARCH and Rule Ensemble. Lastly the future developmentstrategy of Jinjiu Group was suggested.
The yearly population, the disposable personal income and the output of alcohol,peer, wine of China different provinces were organized. The Spearman and Kendallrank correlation coeffcients were calculated for each two of them. The result showedthat peer, wine, population and income all have in?uence on alcohol. Therefore, alcoholoutput was considered as dependent variable and the other four indicators as dependencevariables. Their function was fitted by Random Forest and the in?uence strength wasmake out. Partial dependence was plotted.
Based on monthly profit of Jinjiu Group from January, 1986 to December, 2006,stationary and ARCH tests manifested that mean of one lag di?erence can be modeledby linear ARMA and variance by GARCH. ARMA(1,14)-GARCH(1,1) was determinedby testing the independence of the residuals. Parameters were estimated by maximumlikelihood function, then future profit was demonstrated.
Monthly output, income and tax of the same time of profit in Jinjiu Group wereprepared. Taking profit as output variable and the other three as input variable, theirfunction was fitted by Rule Ensemble. Input variable was predicted by GARCH. Profitwas predicted by substituting predicted input variable to fitted function. Testing resultdemonstrated that Rule Ensemble is superior to GARCH by real profit from January,2007 to June, 2007.
Present condition, market character and developing trend of alcohol industry inChina and Tianjin were analyzed. Based on influence strength between alcohol andbeer and wine and predicted profit, the chance, challenge and diffculty of Jinjiu Groupwere explained. Its developing objects, idea and plan were suggested from constructingthree platforms, increasing six kinds of ability, integrating administrate structure andgrouping management mode. All were very useful for Jinjiu Group’s expanding.
根据全国各省市历年白酒、啤酒、葡萄酒的产量与各省市人口及城镇居民人均可支配收入数据，计算以上数据两两之间的Spearman和Kendall秩相关系数，结果表明啤酒、葡萄酒、人口和收入都影响着白酒产量。因此将白酒产量当作因变量，其它4个指标当作自变量，首次利用随机森林法拟合它们之间的函数关系，然后基于此拟合函数计算其它指标对于白酒产量的影响程度，并做出了偏相关图形。
通过津酒集团1986年1月到2006年12月20年每月的利润数据，首先根据数据的平稳性和ARCH效应检验结果，首次确定利润一阶差分均值采用线性ARMA模型，方差采用广义条件异方差GARCH模型。再通过检验残差的独立性，最后确定出ARMA(1,14)-GARCH(1,1)模型，极大似然估计参数，从而预测出津酒集团未来利润。
按照津酒集团利润数据和相同时段1986年1月到2006年12月每月产量、收入和税金。把利润当成因变量，产量、收入和税金当做自变量，首次利用规则集成法拟合它们之间的函数关系。采用GARCH模型预测自变量――产量、收入和税金未来对应数值，将预测出的自变量数值再代入拟合函数，从而预测未来利润。并用2007年1至6月的利润值进行检验，结果表明规则集成法预测误差更小。
通过全国和天津白酒行业现状、市场特征和发展趋势的分析，根据本文研究的啤酒、葡萄酒及白酒产量的影响程度和白酒利润预测结果，指出天津津酒集团面临机遇、挑战和发展瓶颈，提出津酒集团发展目标、思路和战略规划：从构建三大平台、提升六大能力、整合公司治理结构与集团化管理模式等方面的实施方案，从而为津酒集团的战略腾飞指明方向。
Liquor product is traditional industry in China. It plays an important role in na-tional economic development. Market and profit are the base for its surviving and devel-oping. The dependence between liquor output and their influential factors was analyzedby rank correlation coeffcient and Random Forest. The profit of Tianjin Jinjiu Groupwas predicted by ARMA-GARCH and Rule Ensemble. Lastly the future developmentstrategy of Jinjiu Group was suggested.
The yearly population, the disposable personal income and the output of alcohol,peer, wine of China different provinces were organized. The Spearman and Kendallrank correlation coeffcients were calculated for each two of them. The result showedthat peer, wine, population and income all have in?uence on alcohol. Therefore, alcoholoutput was considered as dependent variable and the other four indicators as dependencevariables. Their function was fitted by Random Forest and the in?uence strength wasmake out. Partial dependence was plotted.
Based on monthly profit of Jinjiu Group from January, 1986 to December, 2006,stationary and ARCH tests manifested that mean of one lag di?erence can be modeledby linear ARMA and variance by GARCH. ARMA(1,14)-GARCH(1,1) was determinedby testing the independence of the residuals. Parameters were estimated by maximumlikelihood function, then future profit was demonstrated.
Monthly output, income and tax of the same time of profit in Jinjiu Group wereprepared. Taking profit as output variable and the other three as input variable, theirfunction was fitted by Rule Ensemble. Input variable was predicted by GARCH. Profitwas predicted by substituting predicted input variable to fitted function. Testing resultdemonstrated that Rule Ensemble is superior to GARCH by real profit from January,2007 to June, 2007.
Present condition, market character and developing trend of alcohol industry inChina and Tianjin were analyzed. Based on influence strength between alcohol andbeer and wine and predicted profit, the chance, challenge and diffculty of Jinjiu Groupwere explained. Its developing objects, idea and plan were suggested from constructingthree platforms, increasing six kinds of ability, integrating administrate structure andgrouping management mode. All were very useful for Jinjiu Group’s expanding.
引文
[1] Hamilton J D. Time Series Analysis[M]. New Jersey: Princeton University Press,1994.
[2] Kantz H, Schreiber T. Nonlinear Time Series Analysis: Second Edition[M].London: Cambridge University Press, 2006.
[3] Engle R. Autorregressive Conditional Heteroskedasticity with Estimates ofUnited Kingdom In?ation[J]. Econometrica, 1982, 50:987–1008.
[4] Bollerslev T. Generalized Autorregressive Conditional Heteroskedasticity[J].Journal of Econometrics, 1986, 31:307–327.
[5] Gurney K. An Introduction to Neural Networks[M]. London: UCL Press, 1997.
[6] Friedman J H, Stuetzle W. Projection pursuit regression[J]. Journal of The Amer-ican Statistical Association, 1981, 76:817–823.
[7] Fredman J. Multivariate adaptive regression splines(with discussion)[J]. Theannals of statistics, 1991, 19:1–141.
[8] Vapnik V. The Nature of Statistical Learning Theory[M]. New York: Springer,1995.
[9] Freidman J H. Greedy function approximation: a gradient boosting machine[J].Annals of Statistics, 2001, 29:1189–1232.
[10] Hastie T, Tibshirani R. Genaralized Additive Models[M]. London: Chapmanand Hall, 1990.
[11] Breiman L, Fredman J, Olshen R, et al. Classification and Regression Trees[M].New York: Chapman and Hall, 1984.
[12] Breiman L. Random Forests[J]. Mach. Learn, 2001, 45(1):5–32.
[13] Jiang W. Some Theoretical Aspects of Boosting in the Presence of NoisyData[C]//Proc. 18th International Conf. on Machine Learning.[S.l.]: MorganKaufmann, San Francisco, CA, 2001:234–241.
[14] Lawrence R, Bunn A, Powell S, et al. Classification of remotely sensed imageryusing stochastic gradient boosting as a refinement of classification tree analy-sis[J]. Remote Sensing of Environment, 2004, 90(3):331–336.
[15] Hancock T, Put R, Coomans D, et al. A performance comparison of modernstatistical techniques for molecular descriptor selection and retention predictionin chromatographic QSRR studies[J]. Chemometrics and Intelligent LaboratorySystems, 2005, 76(2):185–196.
[16] Bricklemyer R S, Lawrence R L, Miller P R, et al. Predicting tillage practices andagricultural soil disturbance in north central Montana with Landsat imagery[J].Agriculture, Ecosystems & Environment, 2006, 114(2-4):210–216.
[17] Moisen G G, Freeman E A, Blackard J A, et al. Predicting tree species presenceand basal area in Utah: A comparison of stochastic gradient boosting, generalizedadditive models, and tree-based methods[J]. Ecological Modelling, 2006,199(2):176–187.
[18] Bricklemyer R S, Lawrence R L, Miller P R, et al. Monitoring and verifying agriculturalpractices related to soil carbon sequestration with satellite imagery[J].Agriculture, Ecosystems & Environment, 2007, 118(1-4):201–210.
[19]韩永奇.我国啤酒市场半年分析及其未来走向预测[J].啤酒科技,2006(10):6–7.
[20]红叶.专家预测酒业未来走势[J].中国食品, 2005(Z1):12–13.
[21]周世水.中国啤酒发展的预测模型及分析[J].酿酒, 2005, 32(1):8–10.
[22]杨君岐,邢战雷,李青,等.啤酒市场的季节规律分析、趋势预测及数据挖掘算法[J].酿酒科技, 2004(6):104–107.
[23]王宝鉴,黄玉霞,许东蓓,等.浅析啤酒销售与气象要素的关系及预测[J].甘肃气象, 2003, 21(2):25–27.
[24]乔淑红,孙德宝.对啤酒行业中销售预测的分析与研究[J].啤酒科技,2000(11):54–58.
[25] ThodeJr H C. Testing for Normality[M]. New York.: Marcel Dekker, 2002.
[26]史道济.相关系数与相关性[J].统计科学与实践, 2002, 4:22–24.
[27] Embrechts L F M A, P. Handbook of Heavy Tailed Distributions in Finance[M].Rachev: Elsevier, 2003: 329–384.
[28] Nguyen H N, Ohn S Y. DRFE: Dynamic Recursive Feature Elimination for geneidentification based on Random Forest[J]. Neural Information Processing, Pt 3,Proceedings, 2006, 4234:1–10.
[29] Enot D P, Beckmann M, Draper J. On the interpretation of high throughput MSbased metabolomics fingerprints with Random Forest[J]. Computational LifeSciences Ii, Proceedings, 2006, 4216:226–235.
[30] Barcellos L F, Ramsay P P, Madden E, et al. Random forest analysis: a novel approachfor exploring the complex genetic component in rheumatoid arthritis.[J].Arthritis And Rheumatism, 2006, 54(9):S132–S132.
[31] Palmer D S, O’Boyle N M, Glen R C, et al. Random forest models to predictaqueous solubility[J]. Journal Of Chemical Information And Modeling, 2007,47(1):150–158.
[32] Zhang Q Y, Aires-De-Sousa J. Random forest prediction of mutagenicity fromempirical physicochemical descriptors[J]. Journal Of Chemical Information AndModeling, 2007, 47(1):1–8.
[33] Strobl C, Boulesteix A L, Zeileis A, et al. Bias in random forest variable importancemeasures: Illustrations, sources and a solution[J]. Bmc Bioinformatics,2007, 8:25.
[34] Ehrman T M, Barlow D J, Hylands P J. Virtual screening of Chinese herbswith random forest[J]. Journal Of Chemical Information And Modeling, 2007,47(2):264–278.
[35] Granitto P M, Gasperi F, Biasioli F, et al. Modern data mining tools in descriptivesensory analysis: A case study with a Random forest approach[J]. Food QualityAnd Preference, 2007, 18(4):681–689.
[36] Bayly C I, Brideau C, Liaw A, et al. Iterative focussed screening using RandomForest: A comparison with HTS/random screening for two extreme cases[J]. AbstractsOf Papers Of The American Chemical Society, 2006, 231.
[37] Granitto P M, Furlanello C, Biasioli F, et al. Recursive feature elimination withrandom forest for PTR-MS analysis of agroindustrial products[J]. ChemometricsAnd Intelligent Laboratory Systems, 2006, 83(2):83–90.
[38] Efron B. Bootstrap Methods: Another Look at the Jackknife[J]. The Annals ofStatistics, 1979, 7(1):1–26.
[39] Singh K. On the Asymptotic Accuracy of Efron’s Bootstrap[J]. The Annals ofStatistics, 1981, 9(6):1187–1195.
[40] Liaw A, Wiener M. Classification and regression by randomForest[J]. Rnews,2002, 2:18–22.
[41]王春峰,蒋祥林,吴晓霖.随机波动性模型的比较分析[J].系统工程学报,2005, 20(2):216–219.
[42]王春峰,程鹏飞,房振明.中国股市流动性与波动性关系的实证研究[J].北京理工大学学报：社会科学版, 2007, 9(4):63–66.
[43]王春峰,韩冬,蒋祥林.基于信息非对称模型的交易行为与波动性关系研究―交易规模和交易频率[J].管理工程学报, 2007, 21(1):134–137.
[44]王春峰.基于时间特性的中国股市交易集群性特征的研究[J].管理工程学报, 2006, 20(2):28–33.
[45]房振明,王春峰.交易频率视角下我国股市量价关系研究[J].系统工程学报,2007, 22(4):386–393.
[46] Mandelbrot B. The variation of certain speculative prices[J]. Journal of Business,1963, 36:394–419.
[47] Mandelbrot B, Taylo H. On the distribution of stock price differences[J]. OperationsResearch, 1967, 15:1057–1062.
[48]罗登跃,王春峰.上证指数收益率、波动性与成交量动态关系研究―基于日数据的非线性动力学实证分析[J].系统工程理论与实践, 2005, 25(7):41–48.
[49] Bali R, Guirguis H. Extreme observations and non-normality in ARCH andGARCH[J]. International Review of Economics & Finance, 2007, 16(3):332–346.
[50] Bellini F, Bottolo L. Stationarity domains for [delta]-power Garch process withheavy tails[J]. Statistics & Probability Letters, 2007, 77(13):1418–1427.
[51] Cook S. The impact of GARCH on asymmetric unit root tests[J]. Physica A:Statistical Mechanics and its Applications, 2006, 369(2):745–752.
[52] Christian F, Jean-Michel Z. Deriving the autocovariances of powers of MarkovswitchingGARCH models, with applications to statistical inference[J]. ComputationalStatistics & Data Analysis, 2007, In Press, Corrected Proof:–873.
[53] Gerlach R, Tuyl F. MCMC methods for comparing stochastic volatility andGARCH models[J]. International Journal of Forecasting, 2006, 22(1):91–107.
[54] Hang Chan N, Deng S J, Peng L, et al. Interval estimation of value-at-risk basedon GARCH models with heavy-tailed innovations[J]. Journal of Econometrics,2007, 137(2):556–576.
[55] Jondeau E, Rockinger M. The Copula-GARCH model of conditional dependencies:An international stock market application[J]. Journal of InternationalMoney and Finance, 2006, 25(5):827–853.
[56] Ling S. Self-weighted and local quasi-maximum likelihood estimatorsfor ARMA-GARCH/IGARCH models[J]. Journal of Econometrics, 2007,140(2):849–873.
[57] Liu J C. Stationarity for a Markov-switching Box-Cox transformed thresholdGARCH process[J]. Statistics & Probability Letters, 2007, 77(13):1428–1438.
[58] Pascual L, Romo J, Ruiz E. Bootstrap prediction for returns and volatilitiesin GARCH models[J]. Computational Statistics & Data Analysis, 2006,50(9):2293–2312.
[59] Sohn S Y, Lim M. Hierarchical forecasting based on AR-GARCH modelin a coherent structure[J]. European Journal of Operational Research, 2007,176(2):1033–1040.
[60] Wen Cheong C, Hassan Shaari Mohd Nor A, Isa Z. Asymmetry and longmemoryvolatility: Some empirical evidence using GARCH[J]. Physica A: Statisticaland Theoretical Physics, 2007, 373:651–664.
[61] deWetW. A structural GARCH model: An application on South African data[J].Economic Modelling, 2006, 23(5):775–791.
[62] Wong K F K, Galka A, Yamashita O, et al. Modelling non-stationary variancein EEG time series by state space GARCH model[J]. Computers in Biology andMedicine, 2006, 36(12):1327–1335.
[63] Yang J P, Cheng S H, Zhang L H. Bivariate copula decomposition in terms ofcomonotonicity, countermonotonicity and independence[J]. Insurance Mathematics& Economics, 2006, 39(2):267–284.
[64]王春峰,张蕊,房振明,等.高频数据下投资组合风险预测模型比较[J].系统工程, 2007, 25(3):23–28.
[65]刘志东,徐淼.基于GARCH和EVT的金融资产风险价值度量方法[J].统计与决策, 2007(18):13–16.
[66]周颖,仇晓光.基于CVaR―GARCH―GED模型的单品种期货风险价值预测[J].统计与决策, 2007(18):95–97.
[67]苏岩,杨振海. GARCH（1,1）模型及其在汇率条件波动预测中的应用[J].数理统计与管理, 2007, 26(4):615–620.
[68]张跃军,范英,魏一鸣.基于GED-GARCH模型的中国原油价格波动特征研究[J].数理统计与管理, 2007, 26(3):398–406.
[69]杨立洪,徐黄玮,曹显兵.可转换债券风险测度的新方法――GAVaR模型[J].数学的实践与认识, 2007, 37(11):92–98.
[70]刘达,牛东晓,邢棉,等.基于GARCH误差校正的遗传支持向量机日前电价预测[J].电力系统自动化, 2007, 31(11):31–34,58.
[71] Taylor S. Modelling Financial Time Series[M]. New York: Wiley, 1986.
[72] Nelson D B. onditional heteroskedasticity in asset returns: A new approach[J].Econometrica, 1991, 59:347–370.
[73] Schwert G W. Stock Volatility and the Crash of 87[J]. The Review of FinancialStudies, 1990, 3:77–102.
[74] Ding Z, Granger C W J, Engle R F. A long memory property of stock marketreturns and a new model[J]. Journal of Empirical Finance, 1993, 1(1):83–106.
[75] Engle R F, Ng V. Measuring and Testing the Impact of News on Volatility[J].Journal of Finance, 1993, 45:1749–1777.
[76] Glosten L, Jagannathan R, Runkle D. On the Relation Between the ExpectedValue and the Volatility on the Nominal Excess Returns on Stocks[J]. Journal ofFinance, 1993, 48:1779–1801.
[77] Duan J C. Augmented GARCH(p,q) process and its di?usion limit[J]. Journalof Econometrics, 1997, 79(1):97–127.
[78] Milhoj A. The moment structure of ARCH models[J]. Scandinavian Journal ofStatistics, 1985, 12:281–292.
[79] Hentschel L. All in the family Nesting symmetric and asymmetric GARCHmodels[J]. Journal of Financial Economics, 1995, 39(1):71–104. available athttp://ideas.repec.org/a/eee/jfinec/v39y1995i1p71-104.html.
[80] Kon S J. Models of stock returns: a comparison[J]. Journal of Finance, 1984,39:147–165.
[81] Blattberg R, Gonedes N. A comparison of the stable and student distributions asstatistical models for stock prices[J]. Journal of Business, 1974, 47:244–280.
[82] Kwiatkowski D, Phillips P C B, Schmidt P, et al. Testing the Null Hypothesis ofStationarity against the Alternative of a Unit Root[J]. Journal of Econometrics,1992, 54:159–178.
[83] Bollerslev T, Chou R Y, Kroner K F. ARCH modeling in finance : A review of thetheory and empirical evidence[J]. Journal of Econometrics, 1992, 52(1-2):5–59.available at http://ideas.repec.org/a/eee/econom/v52y1992i1-2p5-59.html.
[84] Friedman J H, Popescu B E. Importance Sampled Learning Ensembles[R]. Stan-fard University: Department of Statistics, 2003.
[85] Tibshirani R. Regression shrinkage and selection via the lasso[J]. Journal ofRoyal Statistics Society: Series B, 1996, 58:267–288.
[86] Donoho D, Johnstone I, Kerkyacharian G, et al. Wavelet shrinkage; asymptotia?(with discussion)[J]. Journal of Royal Statistics Society: Series A, 1995, 57:201–337.
[87] Friedman J H, Popescu B E. Gradient directed regularization[R]. Stanford Uni-verstity: Department of Statistics, 2004.
[88] Breiman L. Bagging Predictors[J]. Machine Learning, 1996, 26(2):123.
[89] Friedman J H, Hall P. On Bagging and Nonlinear Estimation[R]. Stanfard Uni-verstiy: Department of Statistics, 1999.
[90] Y F, B. S R. A Decision-Theoretic Generalization of Online Learning and anApplication to Boosting[J]. Journal of Computer and System Sciences, 1997,55(1):119–139.
[91] Friedman J H. Stochastic gradient boosting[J]. Computational Statistics & DataAnalysis, 2002, 38(4):367–378.
[92] Friedman J H, Popescu B E. Predictive learning via rule ensembles[R]. StanfordUniversity: Department of Statistics, 2005.
[93] Canfield R, Pevtsov A. Vector magnetic fields, subsurface stresses and evo-lution of magnetic helicity[J]. Journal of Astrophysics and Astronomy, 2000,21(3):213–220.
[94] Ehrentreich F. Representation of extended infrared spectrum-structure-correlations based on fuzzy theory[J]. Analytical and Bioanalytical Chemistry,1997, 357(5):527–533.
[95] Ehrenstein A, Proctor R W. Selecting mapping rules and responses in mixedcompatibility four-choice tasks[J]. Psychological Research, 1998, 61(4):231–248.
[96] Prasad S N. Simulated Annealing 2-D Nonlinear Inversion of Geomagnetic DeepSounding Data Near Ujjain-Guna, India[J]. Pure and Applied Geophysics, 1999,154(2):343–363.
[97] Huang H, Jiang Z, Yu Z, K.and Lu. A skewed Laplace distribution with financialapplication[J]. Financial Systems Engineering , 2003, 19(2):39–52.
[98] Gri?ths R. Probabilities and Quantum Reality: Are There Correlata?[J]. Foun-dations of Physics, 2003, 33(10):1423–1459.
[99] Hushwater V. A Path from the Quantization of the Action Variable to QuantumMechanical Formalism[J]. Foundations of Physics, 1998, 28(2):167–184.
[100] Gerstner W, Ritz R, van Hemmen J. Why spikes? Hebbian learning and retrievalof time-resolved excitation patterns[J]. Biological Cybernetics, 1993, 69(5):503–515.
[101] Karapiperis T. Cellular automaton model of precipitation/dissolution coupledwith solute transport[J]. Journal of Statistical Physics, 1995, 81(1):165–180.
[102] Valov P, Leiman V. Energy ?uctuations in optical transitions in CuHal crystals[J].Physics of the Solid State, 2002, 44(4):720–726.
[103] Schneider von Deimling T, Held H, Ganopolski A, et al. Climate sensitivityestimated from ensemble simulations of glacial climate[J]. Climate Dynamics,2006, 27(2):149–163.
[104] Wang X, Tang X. Random Sampling for Subspace Face Recognition[J]. Interna-tional Journal of Computer Vision, 2006, 70(1):91–104.
[105] Andreoni A, Puddu E, Bondani M. Parallel spatial intensity correlations to de-code random frequency-downconverted images[J]. Applied Physics B: Lasersand Optics, 2006, 85(1):125–129.
[106] Norinder U, Lidn P, Bostrm H. Discrimination between modes of toxic action ofphenols using rule based methods[J]. Molecular Diversity, 2006, 10(2):207–212.
[107] Zeng A, Pan D, He J B. Prediction of MHC II-binding peptides using roughset-based rule sets ensemble[J]. Applied Intelligence, 2007, 27(2):153–166.
[108] Querlioz D, Dollfus P, Do V N, et al. An improved Wigner Monte-Carlo tech-nique for the self-consistent simulation of RTDs[J]. Journal of ComputationalElectronics, 2006, 5(4):443–446.
[109] Mota J, Esteves I. Simplified gauge-cell method and its application to the study ofcapillary phase transition of propane in carbon nanotubes[J]. Adsorption, 2007,13(1):21–32.
[110] Thiel G, MacRobbie E A C, Blatt M R. Membrane transport in stomatal guardcells: The importance of voltage control[J]. Journal of Membrane Biology, 1992,126(1):1–18.
[111] Gilmore T, Park J. Superselection rules in quantum theory: Part II. Subensembleselection[J]. Foundations of Physics, 1979, 9(9):739–749.
[112] Gilmore T, Park J. Superselection rules in quantum theory: Part I. A new pro-posal for state restriction violation[J]. Foundations of Physics, 1979, 9(7):537–556.
[113] Choquard P, Piller B, Rentsch R. On the dielectric susceptibility of classicalCoulomb systems. II[J]. Journal of Statistical Physics, 1987, 46(3):599–633.
[114] Zabolitzky J G. Critical properties of Rule 22 elementary cellular automata[J].Journal of Statistical Physics, 1988, 50(5):1255–1262.
[115] Pervushin Y V. On the possibility of steady evolution of gas bubbles in a liquid[J].Fluid Dynamics, 1983, 18(6):894–899.
[116] Gerstner W, Ritz R, van Hemmen J L. Why spikes? Hebbian learning andretrieval of time-resolved excitation patterns[J]. Biological Cybernetics, 1993,69(5):503–515.
[117] Hanson J E, Crutchfield J P. The attractor asin portrait of a cellular automaton[J].Journal of Statistical Physics, 1992, 66(5):1415–1462.
[118] Yershov V N. Equilibrium Configurations of Tripolar Charges[J]. Few-BodySystems, 2005, 37(1):79–106.
[119] Hansen K, Falk J. Kinetic energy release during evaporation from large sodiumclusters[J]. Molecules and Clusters, 1995, 34(4):251–256.
[120] Sanderson B, Piani C, Ingram W, et al. Towards constraining climate sensitivityby linear analysis of feedback patterns in thousands of perturbed-physics GCMsimulations[J]. Climate Dynamics, 2008, 30(2):175–190.
[121] Liu J G, Zhang X L, Wu W P. Application of fuzzy neural network for realestate prediction[J]. ADVANCES IN NEURAL NETWORKSISNN 2006, PT 3,PROCEEDINGS, 2006, 3973:1187–1191.
[122] Reformat M, Yager R. Building ensemble classifiers using belief functions andOWA operators[J]. Soft Computing - A Fusion of Foundations, Methodologiesand Applications, 2008, 12(6):543–558.
[123]刁昳.中国啤酒行业竞争态势和企业的动态竞争研究[D].北京:对外经济贸易大学, 2007.
[124]陈永茂.保健酒消费行为研究[D].北京:清华大学, 2004.
[125]黄卜智.燕京啤酒的并购战略评价[D].武汉:华中科技大学, 2006.
[126]张一.衡水老白干酒新品牌营销战略研究[D].天津:天津大学, 2006.
[2] Kantz H, Schreiber T. Nonlinear Time Series Analysis: Second Edition[M].London: Cambridge University Press, 2006.
[3] Engle R. Autorregressive Conditional Heteroskedasticity with Estimates ofUnited Kingdom In?ation[J]. Econometrica, 1982, 50:987–1008.
[4] Bollerslev T. Generalized Autorregressive Conditional Heteroskedasticity[J].Journal of Econometrics, 1986, 31:307–327.
[5] Gurney K. An Introduction to Neural Networks[M]. London: UCL Press, 1997.
[6] Friedman J H, Stuetzle W. Projection pursuit regression[J]. Journal of The Amer-ican Statistical Association, 1981, 76:817–823.
[7] Fredman J. Multivariate adaptive regression splines(with discussion)[J]. Theannals of statistics, 1991, 19:1–141.
[8] Vapnik V. The Nature of Statistical Learning Theory[M]. New York: Springer,1995.
[9] Freidman J H. Greedy function approximation: a gradient boosting machine[J].Annals of Statistics, 2001, 29:1189–1232.
[10] Hastie T, Tibshirani R. Genaralized Additive Models[M]. London: Chapmanand Hall, 1990.
[11] Breiman L, Fredman J, Olshen R, et al. Classification and Regression Trees[M].New York: Chapman and Hall, 1984.
[12] Breiman L. Random Forests[J]. Mach. Learn, 2001, 45(1):5–32.
[13] Jiang W. Some Theoretical Aspects of Boosting in the Presence of NoisyData[C]//Proc. 18th International Conf. on Machine Learning.[S.l.]: MorganKaufmann, San Francisco, CA, 2001:234–241.
[14] Lawrence R, Bunn A, Powell S, et al. Classification of remotely sensed imageryusing stochastic gradient boosting as a refinement of classification tree analy-sis[J]. Remote Sensing of Environment, 2004, 90(3):331–336.
[15] Hancock T, Put R, Coomans D, et al. A performance comparison of modernstatistical techniques for molecular descriptor selection and retention predictionin chromatographic QSRR studies[J]. Chemometrics and Intelligent LaboratorySystems, 2005, 76(2):185–196.
[16] Bricklemyer R S, Lawrence R L, Miller P R, et al. Predicting tillage practices andagricultural soil disturbance in north central Montana with Landsat imagery[J].Agriculture, Ecosystems & Environment, 2006, 114(2-4):210–216.
[17] Moisen G G, Freeman E A, Blackard J A, et al. Predicting tree species presenceand basal area in Utah: A comparison of stochastic gradient boosting, generalizedadditive models, and tree-based methods[J]. Ecological Modelling, 2006,199(2):176–187.
[18] Bricklemyer R S, Lawrence R L, Miller P R, et al. Monitoring and verifying agriculturalpractices related to soil carbon sequestration with satellite imagery[J].Agriculture, Ecosystems & Environment, 2007, 118(1-4):201–210.
[19]韩永奇.我国啤酒市场半年分析及其未来走向预测[J].啤酒科技,2006(10):6–7.
[20]红叶.专家预测酒业未来走势[J].中国食品, 2005(Z1):12–13.
[21]周世水.中国啤酒发展的预测模型及分析[J].酿酒, 2005, 32(1):8–10.
[22]杨君岐,邢战雷,李青,等.啤酒市场的季节规律分析、趋势预测及数据挖掘算法[J].酿酒科技, 2004(6):104–107.
[23]王宝鉴,黄玉霞,许东蓓,等.浅析啤酒销售与气象要素的关系及预测[J].甘肃气象, 2003, 21(2):25–27.
[24]乔淑红,孙德宝.对啤酒行业中销售预测的分析与研究[J].啤酒科技,2000(11):54–58.
[25] ThodeJr H C. Testing for Normality[M]. New York.: Marcel Dekker, 2002.
[26]史道济.相关系数与相关性[J].统计科学与实践, 2002, 4:22–24.
[27] Embrechts L F M A, P. Handbook of Heavy Tailed Distributions in Finance[M].Rachev: Elsevier, 2003: 329–384.
[28] Nguyen H N, Ohn S Y. DRFE: Dynamic Recursive Feature Elimination for geneidentification based on Random Forest[J]. Neural Information Processing, Pt 3,Proceedings, 2006, 4234:1–10.
[29] Enot D P, Beckmann M, Draper J. On the interpretation of high throughput MSbased metabolomics fingerprints with Random Forest[J]. Computational LifeSciences Ii, Proceedings, 2006, 4216:226–235.
[30] Barcellos L F, Ramsay P P, Madden E, et al. Random forest analysis: a novel approachfor exploring the complex genetic component in rheumatoid arthritis.[J].Arthritis And Rheumatism, 2006, 54(9):S132–S132.
[31] Palmer D S, O’Boyle N M, Glen R C, et al. Random forest models to predictaqueous solubility[J]. Journal Of Chemical Information And Modeling, 2007,47(1):150–158.
[32] Zhang Q Y, Aires-De-Sousa J. Random forest prediction of mutagenicity fromempirical physicochemical descriptors[J]. Journal Of Chemical Information AndModeling, 2007, 47(1):1–8.
[33] Strobl C, Boulesteix A L, Zeileis A, et al. Bias in random forest variable importancemeasures: Illustrations, sources and a solution[J]. Bmc Bioinformatics,2007, 8:25.
[34] Ehrman T M, Barlow D J, Hylands P J. Virtual screening of Chinese herbswith random forest[J]. Journal Of Chemical Information And Modeling, 2007,47(2):264–278.
[35] Granitto P M, Gasperi F, Biasioli F, et al. Modern data mining tools in descriptivesensory analysis: A case study with a Random forest approach[J]. Food QualityAnd Preference, 2007, 18(4):681–689.
[36] Bayly C I, Brideau C, Liaw A, et al. Iterative focussed screening using RandomForest: A comparison with HTS/random screening for two extreme cases[J]. AbstractsOf Papers Of The American Chemical Society, 2006, 231.
[37] Granitto P M, Furlanello C, Biasioli F, et al. Recursive feature elimination withrandom forest for PTR-MS analysis of agroindustrial products[J]. ChemometricsAnd Intelligent Laboratory Systems, 2006, 83(2):83–90.
[38] Efron B. Bootstrap Methods: Another Look at the Jackknife[J]. The Annals ofStatistics, 1979, 7(1):1–26.
[39] Singh K. On the Asymptotic Accuracy of Efron’s Bootstrap[J]. The Annals ofStatistics, 1981, 9(6):1187–1195.
[40] Liaw A, Wiener M. Classification and regression by randomForest[J]. Rnews,2002, 2:18–22.
[41]王春峰,蒋祥林,吴晓霖.随机波动性模型的比较分析[J].系统工程学报,2005, 20(2):216–219.
[42]王春峰,程鹏飞,房振明.中国股市流动性与波动性关系的实证研究[J].北京理工大学学报：社会科学版, 2007, 9(4):63–66.
[43]王春峰,韩冬,蒋祥林.基于信息非对称模型的交易行为与波动性关系研究―交易规模和交易频率[J].管理工程学报, 2007, 21(1):134–137.
[44]王春峰.基于时间特性的中国股市交易集群性特征的研究[J].管理工程学报, 2006, 20(2):28–33.
[45]房振明,王春峰.交易频率视角下我国股市量价关系研究[J].系统工程学报,2007, 22(4):386–393.
[46] Mandelbrot B. The variation of certain speculative prices[J]. Journal of Business,1963, 36:394–419.
[47] Mandelbrot B, Taylo H. On the distribution of stock price differences[J]. OperationsResearch, 1967, 15:1057–1062.
[48]罗登跃,王春峰.上证指数收益率、波动性与成交量动态关系研究―基于日数据的非线性动力学实证分析[J].系统工程理论与实践, 2005, 25(7):41–48.
[49] Bali R, Guirguis H. Extreme observations and non-normality in ARCH andGARCH[J]. International Review of Economics & Finance, 2007, 16(3):332–346.
[50] Bellini F, Bottolo L. Stationarity domains for [delta]-power Garch process withheavy tails[J]. Statistics & Probability Letters, 2007, 77(13):1418–1427.
[51] Cook S. The impact of GARCH on asymmetric unit root tests[J]. Physica A:Statistical Mechanics and its Applications, 2006, 369(2):745–752.
[52] Christian F, Jean-Michel Z. Deriving the autocovariances of powers of MarkovswitchingGARCH models, with applications to statistical inference[J]. ComputationalStatistics & Data Analysis, 2007, In Press, Corrected Proof:–873.
[53] Gerlach R, Tuyl F. MCMC methods for comparing stochastic volatility andGARCH models[J]. International Journal of Forecasting, 2006, 22(1):91–107.
[54] Hang Chan N, Deng S J, Peng L, et al. Interval estimation of value-at-risk basedon GARCH models with heavy-tailed innovations[J]. Journal of Econometrics,2007, 137(2):556–576.
[55] Jondeau E, Rockinger M. The Copula-GARCH model of conditional dependencies:An international stock market application[J]. Journal of InternationalMoney and Finance, 2006, 25(5):827–853.
[56] Ling S. Self-weighted and local quasi-maximum likelihood estimatorsfor ARMA-GARCH/IGARCH models[J]. Journal of Econometrics, 2007,140(2):849–873.
[57] Liu J C. Stationarity for a Markov-switching Box-Cox transformed thresholdGARCH process[J]. Statistics & Probability Letters, 2007, 77(13):1428–1438.
[58] Pascual L, Romo J, Ruiz E. Bootstrap prediction for returns and volatilitiesin GARCH models[J]. Computational Statistics & Data Analysis, 2006,50(9):2293–2312.
[59] Sohn S Y, Lim M. Hierarchical forecasting based on AR-GARCH modelin a coherent structure[J]. European Journal of Operational Research, 2007,176(2):1033–1040.
[60] Wen Cheong C, Hassan Shaari Mohd Nor A, Isa Z. Asymmetry and longmemoryvolatility: Some empirical evidence using GARCH[J]. Physica A: Statisticaland Theoretical Physics, 2007, 373:651–664.
[61] deWetW. A structural GARCH model: An application on South African data[J].Economic Modelling, 2006, 23(5):775–791.
[62] Wong K F K, Galka A, Yamashita O, et al. Modelling non-stationary variancein EEG time series by state space GARCH model[J]. Computers in Biology andMedicine, 2006, 36(12):1327–1335.
[63] Yang J P, Cheng S H, Zhang L H. Bivariate copula decomposition in terms ofcomonotonicity, countermonotonicity and independence[J]. Insurance Mathematics& Economics, 2006, 39(2):267–284.
[64]王春峰,张蕊,房振明,等.高频数据下投资组合风险预测模型比较[J].系统工程, 2007, 25(3):23–28.
[65]刘志东,徐淼.基于GARCH和EVT的金融资产风险价值度量方法[J].统计与决策, 2007(18):13–16.
[66]周颖,仇晓光.基于CVaR―GARCH―GED模型的单品种期货风险价值预测[J].统计与决策, 2007(18):95–97.
[67]苏岩,杨振海. GARCH（1,1）模型及其在汇率条件波动预测中的应用[J].数理统计与管理, 2007, 26(4):615–620.
[68]张跃军,范英,魏一鸣.基于GED-GARCH模型的中国原油价格波动特征研究[J].数理统计与管理, 2007, 26(3):398–406.
[69]杨立洪,徐黄玮,曹显兵.可转换债券风险测度的新方法――GAVaR模型[J].数学的实践与认识, 2007, 37(11):92–98.
[70]刘达,牛东晓,邢棉,等.基于GARCH误差校正的遗传支持向量机日前电价预测[J].电力系统自动化, 2007, 31(11):31–34,58.
[71] Taylor S. Modelling Financial Time Series[M]. New York: Wiley, 1986.
[72] Nelson D B. onditional heteroskedasticity in asset returns: A new approach[J].Econometrica, 1991, 59:347–370.
[73] Schwert G W. Stock Volatility and the Crash of 87[J]. The Review of FinancialStudies, 1990, 3:77–102.
[74] Ding Z, Granger C W J, Engle R F. A long memory property of stock marketreturns and a new model[J]. Journal of Empirical Finance, 1993, 1(1):83–106.
[75] Engle R F, Ng V. Measuring and Testing the Impact of News on Volatility[J].Journal of Finance, 1993, 45:1749–1777.
[76] Glosten L, Jagannathan R, Runkle D. On the Relation Between the ExpectedValue and the Volatility on the Nominal Excess Returns on Stocks[J]. Journal ofFinance, 1993, 48:1779–1801.
[77] Duan J C. Augmented GARCH(p,q) process and its di?usion limit[J]. Journalof Econometrics, 1997, 79(1):97–127.
[78] Milhoj A. The moment structure of ARCH models[J]. Scandinavian Journal ofStatistics, 1985, 12:281–292.
[79] Hentschel L. All in the family Nesting symmetric and asymmetric GARCHmodels[J]. Journal of Financial Economics, 1995, 39(1):71–104. available athttp://ideas.repec.org/a/eee/jfinec/v39y1995i1p71-104.html.
[80] Kon S J. Models of stock returns: a comparison[J]. Journal of Finance, 1984,39:147–165.
[81] Blattberg R, Gonedes N. A comparison of the stable and student distributions asstatistical models for stock prices[J]. Journal of Business, 1974, 47:244–280.
[82] Kwiatkowski D, Phillips P C B, Schmidt P, et al. Testing the Null Hypothesis ofStationarity against the Alternative of a Unit Root[J]. Journal of Econometrics,1992, 54:159–178.
[83] Bollerslev T, Chou R Y, Kroner K F. ARCH modeling in finance : A review of thetheory and empirical evidence[J]. Journal of Econometrics, 1992, 52(1-2):5–59.available at http://ideas.repec.org/a/eee/econom/v52y1992i1-2p5-59.html.
[84] Friedman J H, Popescu B E. Importance Sampled Learning Ensembles[R]. Stan-fard University: Department of Statistics, 2003.
[85] Tibshirani R. Regression shrinkage and selection via the lasso[J]. Journal ofRoyal Statistics Society: Series B, 1996, 58:267–288.
[86] Donoho D, Johnstone I, Kerkyacharian G, et al. Wavelet shrinkage; asymptotia?(with discussion)[J]. Journal of Royal Statistics Society: Series A, 1995, 57:201–337.
[87] Friedman J H, Popescu B E. Gradient directed regularization[R]. Stanford Uni-verstity: Department of Statistics, 2004.
[88] Breiman L. Bagging Predictors[J]. Machine Learning, 1996, 26(2):123.
[89] Friedman J H, Hall P. On Bagging and Nonlinear Estimation[R]. Stanfard Uni-verstiy: Department of Statistics, 1999.
[90] Y F, B. S R. A Decision-Theoretic Generalization of Online Learning and anApplication to Boosting[J]. Journal of Computer and System Sciences, 1997,55(1):119–139.
[91] Friedman J H. Stochastic gradient boosting[J]. Computational Statistics & DataAnalysis, 2002, 38(4):367–378.
[92] Friedman J H, Popescu B E. Predictive learning via rule ensembles[R]. StanfordUniversity: Department of Statistics, 2005.
[93] Canfield R, Pevtsov A. Vector magnetic fields, subsurface stresses and evo-lution of magnetic helicity[J]. Journal of Astrophysics and Astronomy, 2000,21(3):213–220.
[94] Ehrentreich F. Representation of extended infrared spectrum-structure-correlations based on fuzzy theory[J]. Analytical and Bioanalytical Chemistry,1997, 357(5):527–533.
[95] Ehrenstein A, Proctor R W. Selecting mapping rules and responses in mixedcompatibility four-choice tasks[J]. Psychological Research, 1998, 61(4):231–248.
[96] Prasad S N. Simulated Annealing 2-D Nonlinear Inversion of Geomagnetic DeepSounding Data Near Ujjain-Guna, India[J]. Pure and Applied Geophysics, 1999,154(2):343–363.
[97] Huang H, Jiang Z, Yu Z, K.and Lu. A skewed Laplace distribution with financialapplication[J]. Financial Systems Engineering , 2003, 19(2):39–52.
[98] Gri?ths R. Probabilities and Quantum Reality: Are There Correlata?[J]. Foun-dations of Physics, 2003, 33(10):1423–1459.
[99] Hushwater V. A Path from the Quantization of the Action Variable to QuantumMechanical Formalism[J]. Foundations of Physics, 1998, 28(2):167–184.
[100] Gerstner W, Ritz R, van Hemmen J. Why spikes? Hebbian learning and retrievalof time-resolved excitation patterns[J]. Biological Cybernetics, 1993, 69(5):503–515.
[101] Karapiperis T. Cellular automaton model of precipitation/dissolution coupledwith solute transport[J]. Journal of Statistical Physics, 1995, 81(1):165–180.
[102] Valov P, Leiman V. Energy ?uctuations in optical transitions in CuHal crystals[J].Physics of the Solid State, 2002, 44(4):720–726.
[103] Schneider von Deimling T, Held H, Ganopolski A, et al. Climate sensitivityestimated from ensemble simulations of glacial climate[J]. Climate Dynamics,2006, 27(2):149–163.
[104] Wang X, Tang X. Random Sampling for Subspace Face Recognition[J]. Interna-tional Journal of Computer Vision, 2006, 70(1):91–104.
[105] Andreoni A, Puddu E, Bondani M. Parallel spatial intensity correlations to de-code random frequency-downconverted images[J]. Applied Physics B: Lasersand Optics, 2006, 85(1):125–129.
[106] Norinder U, Lidn P, Bostrm H. Discrimination between modes of toxic action ofphenols using rule based methods[J]. Molecular Diversity, 2006, 10(2):207–212.
[107] Zeng A, Pan D, He J B. Prediction of MHC II-binding peptides using roughset-based rule sets ensemble[J]. Applied Intelligence, 2007, 27(2):153–166.
[108] Querlioz D, Dollfus P, Do V N, et al. An improved Wigner Monte-Carlo tech-nique for the self-consistent simulation of RTDs[J]. Journal of ComputationalElectronics, 2006, 5(4):443–446.
[109] Mota J, Esteves I. Simplified gauge-cell method and its application to the study ofcapillary phase transition of propane in carbon nanotubes[J]. Adsorption, 2007,13(1):21–32.
[110] Thiel G, MacRobbie E A C, Blatt M R. Membrane transport in stomatal guardcells: The importance of voltage control[J]. Journal of Membrane Biology, 1992,126(1):1–18.
[111] Gilmore T, Park J. Superselection rules in quantum theory: Part II. Subensembleselection[J]. Foundations of Physics, 1979, 9(9):739–749.
[112] Gilmore T, Park J. Superselection rules in quantum theory: Part I. A new pro-posal for state restriction violation[J]. Foundations of Physics, 1979, 9(7):537–556.
[113] Choquard P, Piller B, Rentsch R. On the dielectric susceptibility of classicalCoulomb systems. II[J]. Journal of Statistical Physics, 1987, 46(3):599–633.
[114] Zabolitzky J G. Critical properties of Rule 22 elementary cellular automata[J].Journal of Statistical Physics, 1988, 50(5):1255–1262.
[115] Pervushin Y V. On the possibility of steady evolution of gas bubbles in a liquid[J].Fluid Dynamics, 1983, 18(6):894–899.
[116] Gerstner W, Ritz R, van Hemmen J L. Why spikes? Hebbian learning andretrieval of time-resolved excitation patterns[J]. Biological Cybernetics, 1993,69(5):503–515.
[117] Hanson J E, Crutchfield J P. The attractor asin portrait of a cellular automaton[J].Journal of Statistical Physics, 1992, 66(5):1415–1462.
[118] Yershov V N. Equilibrium Configurations of Tripolar Charges[J]. Few-BodySystems, 2005, 37(1):79–106.
[119] Hansen K, Falk J. Kinetic energy release during evaporation from large sodiumclusters[J]. Molecules and Clusters, 1995, 34(4):251–256.
[120] Sanderson B, Piani C, Ingram W, et al. Towards constraining climate sensitivityby linear analysis of feedback patterns in thousands of perturbed-physics GCMsimulations[J]. Climate Dynamics, 2008, 30(2):175–190.
[121] Liu J G, Zhang X L, Wu W P. Application of fuzzy neural network for realestate prediction[J]. ADVANCES IN NEURAL NETWORKSISNN 2006, PT 3,PROCEEDINGS, 2006, 3973:1187–1191.
[122] Reformat M, Yager R. Building ensemble classifiers using belief functions andOWA operators[J]. Soft Computing - A Fusion of Foundations, Methodologiesand Applications, 2008, 12(6):543–558.
[123]刁昳.中国啤酒行业竞争态势和企业的动态竞争研究[D].北京:对外经济贸易大学, 2007.
[124]陈永茂.保健酒消费行为研究[D].北京:清华大学, 2004.
[125]黄卜智.燕京啤酒的并购战略评价[D].武汉:华中科技大学, 2006.
[126]张一.衡水老白干酒新品牌营销战略研究[D].天津:天津大学, 2006.