关于带扰动广义Cox保险风险模型的破产概率
关于带扰动广义Cox保险风险模型的破产
概率
第19卷第6期
2007年12月
重庆邮电大学(自然科学版)
JournalofChongqingUniversityofPostsandTelecommunications(NaturalScience) Vo1.19No.6
Dec.2007
Onruinprobabilityforageneralized
Coxinsuranceriskmodelwithperturbation
ZHANDe—Sheng.'!,TANGJia—Shan.
(1.CollegeofProfessi(real_rechnology,AnhuiUniversityofTechnology,Maanshan243011,P.R.China;
2.CollegeofMathematicsandComputerScience,NanjingNormalUniversity,Nanjing210097,P.R.China;
3.CollegeofMathematicsandPhysics.NanSingUniversityofPostsandTelecommunications.Nanjing210003,P.R.
Abstract:Anewinsuranceriskmodelwasinvestigated,inwhichtheinsuranlsarrivalandtheclaimsarrivalaredriv—
enbytwoindependentgeneralizedCoxprocesses.ThemodelisalsoperturbedbyaBrownianmotion.Usingthe
martingalemethod.anupperboundfortheruinprobabilityofthemodelisgiven. Keywords:ruinprobability;upperbound:insuranceriskmodel;martingalemethod CLCnumber:0211.6DocumentCode:AArticleID:1673825X(2007)06—0782—03
关于带扰动广义Cox保险风险模型的破产概率
占德胜,唐加山.
(1.安徽工业大学职,lk技术学院.安徽马鞍山243011;2.南京师范大学数学与计算机
科学学院,江苏南京210097
3南京邮电大学数理学院,江苏南京210003)
摘要:研究了一种新的带布朗运动干扰的保险风险模型.在模型中,投保人以及索
赔都成批到达,到达的点过程
是2个独立的Cox过程.利用鞅方法.给出了该模型破产概率的一个上界.
关键词:破产概率;上界;保险风险模型;鞅方法
1IntrOductiOn
Insurancerisktheoryisoneoftheimportantre—
searchareainbothoftheactuarialscienceandthe appliedprobability.Ruinprobahilityforriskrood—
elsisoneofthehottopicsintherisktheoryduring thepastdecades.FortheclassicalCramer—Iund—
bergmodel,theriskprocessorthesurplusprocess iswrittenby
U()一"+一?Y(1)
where"?0representstheinitialcapitalofthein—
surer,c>0thepremiUIllincomerate,N()the claimnumberprocesswhichismodeledbyahomo—
geneousPoissonprocessandYtheclaimsizesatis—
fyingthatY,_i一1,2,…arei.i.d.nonnegative
randomvariables.
Undercertainconditions.theruinprobabilities andmanyotherresultsoftherelatedproblemsfor thiskindofmodelareobtainedr1.2].However, fromthepracticalpointofview,thisclassicalrisk modelisunrealistic.Therefore,variantextensions oftheclassicalmodelhavebeenpresentedinthe literatureandthenewonesaremorerealistically motivated[1].Specifically.ifthepremiumprocess
*Receiveddate:2007—03—0l
ischangedfromf,thesecondtermintheright M(t)
handsideofequation(1),to?Y,,whereM(f)is
ahomogeneousPoissonprocessandYisthepre—
miumchargedfromthei-thpolicyholder,i一1,2,
…
,thenthenewriskmodeIiScalleddoublePois—
soninsuranceriskmodel(seee.g.[3]).Ifthein—
surantarrivingprocessM(t)andtheclaimnumber processN(t)arechangedtomoregeneralproces—
ses,Coxprocesses,thenthenewmodeliscalled doubleCoxinsuranceriskmodel(forexamplesee J4,5).Besidestheextensionofthepoint process,riskmodelswithinterferencearealsoin—
vestigatedinrecentyears(seee.g.[6]). Inallthemodelsmentionedabove,ifthereis oneunfortunatethinghappened,thenthereisonly oneinsurantwhowillclaimforalOSScoveredby thepolicy.However,inpracticalsituation,forin—
stancetheiDSUranceforvehicles,ifonetrafficacci—
denthappens.thentherearealwaysatIeasttwo insurantswhowillapplyforthecompensationfor thelosses[7,8].Motivatedbythispracticalap- plication,inthisnote,weinvestigateanewmod—
el.inwhichthearrivingprocessoftheinsurants andtheclaimnumberprocessaremodeledbytwo
第6期ZHANDeSt1eng?eta1:()nrLfinprobabilityforageneralizedCoxinsuranceriskmodelwitt1
I)erfurbation?783.
independentgeneralizedCoxprocessesandthe modelisalsoperturbedbyaBrownianmotion.By usingthemartingalemethod,wegiveanupper boundoftheruinprobabilityofthemode1. 2Mainresultsanditsproofs
Inthissection,wegivethemainresultsandits proofsofthispaper.
2.1Modeldescription
Ietl(f),f?0}and!(),f?0}betwoinde
pendentnonnegativestochasticprocesses,and l(f),f?0and!(),f?0}betwoindependent
standardPoissonprocesses,whichareindependent from,(f),?0},=1,2.Then(f)一
,?(A(t))isaCoxprocesswiththeintensity processgivenby(f),whereA()=i()dsis theaccumulatedintensityfunction,一1,2.Iet
{/,2(Z),Z?1}(i一1,2)betwoindependentstochas—
ticseriesof..d.randomvariableswiththecom—
mondistributionsgivenbyP(/,2l(?)一J)一P,and
P(2(?)一)一q,,J一1,2,…,Then
?…
N,(f)一/,2i(z)
,一】
isageneralizedhomogeneousCoxprocesses,一1,2.
Iet一{,,?1}andy一{y,?1}betwose
riesofi.i.d.randomvariables,thentheinsurance riskmodelinvestigatedinthispaperisdefinedas follows
(,)N2(,
u(f)一"+?,一?y,+(f)(2)
Where>0isaconstantandB(f),f?0}isa
standardBrownianmotion.Supposethatallsto—
chasticprocessesandrandomseriesareindepend—
entfromeachother.Forthesurvivaloftheinsur—
?1Nn
ancecompany,wesupposethatE(?,,?y,)
,一』J—J
>0.
Theintuitivemeaningofourriskmodelisasfol lows:"representstheinitialcapitaloftheinsurer, N()thenumberofpointatwhichthereisonein—
surantorabathofinsurantsarriveduringthetime intervalr0,f].Thenumberofinsurantsatthel-th pointisdenotedby/,2l(Z).isthepremium chargedfromtheJthinsurant.N2(f)standsfor thenumberofpointatwhichthereisoneorabath ofclaimingoccurringduringthetimeperiod[0,f]. Thenumberofclaimingatthelthpointisdenotedby /,22(Z)andy_istheclaimsizeoftheJthclaim.The perturbationitemaB(f)representstheuncertaingain and/orlossoftheinsurancecompany.
2.2Ruinprobability
Definetheruintimeoftheriskmodelas:
T:一infU(f)<0U(0)一"}
f?()
Then(")一P(T<IU(0)一")istheruinprob
abilityoftheinsuranceriskmode1.Forthec.n venienceinwhatfollows,weintroducethefollow ingnotation.
以I
一以1(),f?0},
VA2
.
歹一以2(f),?0},
{Nl(),?f},
一
{N!(),?f},一B(),?f},
if一nVV2V
Ifdefine
D:一X(卜l1_l+…+XZ—l,2,…
Q/:一卜1)..1+…+X?,Z一1,2,…(3)
where/,2(0)一0(i=1,2),thenitiseasytoprove that{Df,Z?1}and{Q,Z?1}aretwoindependent
seriesesofi.i.d.randomvariables,withthecorn mortdistributionsgivenbyF『j()一?PF(')
一1
andFQ(z)一?qFi'-Cr),for?0,whereF(,)
isthevalueatofthedistributionfunctionofthe lfoldconvolutionofthedistributionoftherandom variableZ.Furthermore,similartoIemma2.2in [9],wehavethefollowinglemma.
LemmalFortherandomvariablesdefinedin formula(3),theinsuranceriskmodeldefinedin formula(2)isequivalenttothefollowing '{,lt,{l?
"+?D一?Q+(f)(4)1,1
Beforestatingthemainresult,wegiveakey lemma.Forthesake,forr?0,wedenoteand
supposetheyexist,
h】(r)一le…Ff)(d_1,)一1,
J0
h2(r)=
Lemma2Ifwe
M
:
Ie…FQ(d)一1J0
define
exp(一())
1'
exp(hl(r)以l()+!(t)A2(f)+Tt)
thenM,?0}isamartingale.
ProofTheproofisastandardoneandissimi lartothatintheliterature.Wegiveasketchhere. N,
Firstlv,wepr()ve1hatE(e/l.,.I)一eA1(t)hI',
,2?
thenprovethatE(e,jI!)一eAz(tHQ'.Fina1
ly,usingtheconditionalexpectationandproperties
oftheBrwonianmotion,provethemartingale property,i.e.E(M)一Mfor0??f.This
completestheproof.
Hereisthemajnresl】1t
?
784?重庆邮电大学(自然科学皈)第19卷
Theorem1Fortheinsuranceriskmodelde{lrle informula(4).theruinprobabilityininfinitytime
hasanupperboundgivenby(")?(11(7)?e一,
where
c(r)一E(exp(矗(,)n()^:(r)^()斗—)).
ProofBasedontheresultobtainedinIemma 2.theproofofthetheoremissimilartothatof Theorem2.2inl10.Thesketchoftheproofisas thefollowing.
Foranyt<,itcanbeshowi]thatexp(7") 一
M.?E.(MT^,I丁?t)P(丁?t.).Itimplies
from(丁)<OthatE,(M7了?t)?infexp(一
(l(r)AI(t)一h2(r)A!(t)一去f7一)).Therefore,
P(丁?t0)?exp(r")SuI)exp(hI(7)A](f),
?lf
1h
,2(r)A,2(f)+?toy:r).Takingexpectationonboth gettheresult.The
Itiseasytoseetllatforaconcrete,一.(,(r)de—
finedinTheorem3isaconstantan(1thedecayrate oftheupperboundise一".Apparently.1arger,
resultsinmorepreciousasymptoticbehavior.If define
R—sup{r,('(7')一一,
similarlytothediscussionafterTheorcm2.2in L1O,wehaveasimilarcoro1lary.
CorollarylForany0??R.tileruinproba—
bility(")forthegeneralizedCoxinsurancerisk modelhasthefollowingupperbound.
(")?C(R一?)e.
Remark1Theconcreteformoftheresultsob—
rainedinthisnoteissireilartothoseintIlelitera—
ture(seeTheorem2.2an(1Corollary2.3inr10),
however,thecontentiverydifferent.Specifical—
lY,asanexample,(,一)define(1inf()rmula(5)is
closelyrelatedtoboththesizeoftl1e1)atchclai—
mingandthelOSIclaime(1bytheifiSHrants.wl1ilea similarfunctionh(r)iI1r10]isonlydepex1denton theclajms;zeoftl1ensl1rFli1ts.
3Concludingremarks
Inthisnote.weinvesligateanewinSllrancerisk model,inwhicbtheinsurantsarriv,1]andt1]e claimsarrivalaredriYenbvtwoindependentCox processes.Ateachofinsurallts}lrrivalpoint,there isabatchofinstlran1.swhopaythepremiuman(1at eachofelajHiSpoint.thereis}lbgItchofiIlSt1l-a1I1s whoclai12]theJOSt..per',urbatjollofBrwoRiaH motionisalsointrodtlced.Usingtilemartingale method,wegiveanupperboundoftheruinproba—
bilityfortheriskmode1.
References:
:1]
:2]
:3]
10
L,HENGShixue.I,he5urveyforresearcheso{ruin theoryJ(inChinese).AdvancesinMathematics (China).2002,31(5):403—422.
EEBRECHTSP.KIUPPEIBERGC,MIK0SCHT.
Modellingextremateventsforinsuranceandfinance —M].Bcrlin:SpringerVerlag,1997.
ITUBaoliang.WANGYong—mao.WENYan,qing.
RuinprobabilityindoublePoissonriskmodel[J](in
Chinese).Journa1ofYanshanUniversity.2006,3O (4):296-299.
ZE(Ai—lln,JJJNXiang,ZHANGHanjun.ACox
riskmodelofdoubleline[J~(inChinese).Mathemati—
calTheoryandApplications,2003,23(1):107-112. HEIi-na.IIUZai—ming.Thediscussionofaspecial CoxriskmodelIJ](inChinese).J.ofGuangxiUni一
,ersityforNationalities.2006,12(2):8O一82.
IU()Jian一}lUa,FAN(Shi—zu.Ruinprobabilityinrisk modelwithgeneralizeddoublePoissonprocessper—
turbedb3rdiffusionJ](inChinese).Mathematical FheoryandApplications.2006,26(3):102—104.
HE.Qixiang.Theprobabilityofruinwithmultiple iIlsuraneeproducts[J](inChinese).MathematicaIthe—
orvandapplications,2005,25(3):68—71.
ZHAN(Kui.Theriskmode}fordouble—insurance
wilhbatchpermiunlsr(inChinese).Mathematical Theor,
yandApplications.2006,26(1):53—56.
YUWen一ang.Theprobabilityofruinmultitype-in—
suranceriskmode}withcompoundgeneralizedhomo—
geneousPoissonprocess[J~(inChinese).Comm.On .
Xpp1.Math.AndComput.2003,17(3):63—69.
XIANGYang.RuinprobabilitiesinaCoxriskmodel perturbedbybrownianmorionJ](inChinese).Jour—
nalofHuaihuaUniversilv.2005,24(2):3—5.
作者简介
占德胜(I968一).男,安徽省怀宁人,讲师,南
京师范大学硕士研究生,主要研究方向为 偏微分方程:E—mail:zhandesheng559@
163.eoHl.
唐加山(1968一).男,安徽天长人,教授,博 士.硕士生导师,主要研究方向为应用概
率,随机过程,排队论以及信号与信息处理 等,发
论文3o余篇.E—mail:jiashant@ yahoo.ca.
w
g一
dC