第34卷第10期系统工程理论与实践V01.34,No.102014年10月SystemsEngineeringTheory&PracticeOct.,2014==========================================================;========================================================================================================文章编号:1000-6788(2014)10-2465-18中图分类号:F830文献标志码:A基于参数学习的GARCH动态无穷活动率Levy过程的欧式期权定价吴恒煜-二,朱福敏,,s,胡根华.,温金明t(1.西南财经大学金融安全协同创新中心经济信息工程学院,成都611130;2.西南财经大学中国金融研究中心,成都611130;3.纽约州立大学石溪分校应用数学与统计系商学院,纽约11794;4.加拿大麦吉尔大学数学与统计学院,蒙特利尔H3A2K6)摘要在股票价格中引入漂移率、波动率和随机跳跃三种状态,建立动态状态空间模型,并通过局部风险中性定价关系(RNVR)推导无套利定价模型,以非高斯条件ARMA-NGARCH为基准模型,构建S&P500指数的离散动态Levy过程,并基于序贯贝叶斯的参数学习方法,进行模型估计和期权定价研究.结果表明:动态Levy过程能够联合刻画时变漂移率、条件波动率和无穷活动率等特征,且贝叶斯方法的引入提高了期权隐含波动率的定价精度,同时,无穷活动率模型在期权定价方面具有显著优势.在五类滤波中,无损粒子滤波估计精度最高,速降调和稳态过程(RDTS)的期权定价误差最小,而非高斯模型在收益率预测方面没有表现出显著的差异.关键词动态Levy过程;杠杆效应;粒子滤波;参数学习;期权定价EuropeanoptionpricingforGARCHdynamicinfiniteactivityLevyprocessesbasedonparameterlearningWUHeng-yu'.2,ZHUFu_min'.3,HUGen-hua',WENJin-ming4(1.CollaborativeInnovationCenterofFinancialSecurity,SchoolofEconomicInformationEngineering,SouthwesternUniversityofFinance&Economics,Chengdu611130,China;2.CenterofChineseFinancialStudies,SouthwesternUniversityofFinance&Economics,Chengdu611130,China;3.DepartmentofAppliedMathematics&Statistics,CollegeofBusiness,SUNYatStonyBrook,NewYork11794,USA;4.DepartmentofMathematics&Statistics,McGillUniversity,MontrealH3A2K6,Canada)AbstractInthispaper,weconsiderathree-dimensionstatespacemodelforestablishingadiscrete-timedynamicLevyprocess,includingtime-varyingdrift,conditionalvolatilityandstochasticjumpactivity.Thenweobtaintheequivalentnon-arbitragepricingmodelthroughlocalrisk-neutralvaluationrelationship(RNVR).Takingnon-GaussianARMA-NGARCHmodelasourbenchmark,weconstructadiscretetimedynamicLevyprocesswithGARCHeffectformodelingS&P500index.FurthermorewejointlyestimatetheparametersofthemodelandstudytheoptionpricingperformancebasedonBayesianlearningapproach.ResearchresultsshowthatourdynamicLevyprocesscandepictthetime-varyingdriftrate,conditionalvolatilityandinfiniteactivitystyles.Meanwhile,Bayesianapproachimprovestheoptionvaluationabilityofourmodel.Infinitejumpmodelsaresignificantsuperiorandincreasethepricingaccuracyofimpliedvolatility.Wealsofindthatunscentedparticlefiltering(UPF)hasthebestestimationperformance,non-Gaussianmodelsintheyieldpredictionareofnosignificantdifference.buttherapidlydecreasingtemperedstableprocesses(RDTS)havenummumerrorsforoptionpricing.KeywordsdynamicLevyprocess:leverageeffects;particLefiltering;parameterlearning;optionpricing收稿日期:2013-03-01资助项目:国家自然科学基金重大研究计划(91218301);国家自然科学基金面上项目(71171168);教育部人文社会科学重点研究基地重大项目(12JJD790026);国家教育部留学基金(201206980001);西南财经大学中央高校基本科研业务费专项资金(JBK1407164,JBK12050);中央高校科研业务费专项资金及四川省教育厅创新团队项目(JBK130401)作者简介:吴恒煜(1970-),男,广东雷州人,博士后,教授...