資源描述：
《FOR FINANCIAL TIME-SERIES FORECASTING》由會員上傳分享，免費在線閱讀，更多相關(guān)內(nèi)容在行業(yè)資料-天天文庫。
1、ALONGMEMORYPATTERNMODELLINGANDRECOGNITIONSYSTEMFORFINANCIALTIME-SERIESFORECASTINGSameerSingh{s.singh@exeter.ac.uk}UniversityofExeterDepartmentofComputerSciencePrinceofWalesRoadExeterEX44PTSingh,S."ALongMemoryPatternModellingandRecognitionSystemforFinancialForecasting",PatternAnalysi
2、sandApplications,vol.2,issue3,pp.264-273,1999.1ALONGMEMORYPATTERNRECOGNITIONANDMODELLINGSYSTEMFORFINANCIALTIME-SERIESFORECASTINGABSTRACTInthispaper,theconceptofalongmemorysystemforforecastingisdeveloped.PatternModellingandRecognitionSystemsareintroducedaslocalapproximationtoolsforfo
3、recasting.Suchsystemsareusedformatchingcurrentstateofthetime-serieswithpaststatestomakeaforecast.Inthepast,thissystemhasbeensuccessfullyusedforforecastingtheSantaFecompetitiondata.Inthispaper,weforecastthefinancialindicesofsixdifferentcountriesandcomparetheresultswithneuralnetworkso
4、nfivedifferenterrormeasures.Theresultsshowthatpatternrecognitionbasedapproachesintime-seriesforecastingarehighlyaccurateandtheseareabletomatchtheperformanceofadvancedmethodssuchasneuralnetworks.21.MOTIVATIONTime-seriesforecastingisanimportantresearchareainseveraldomains.Traditionall
5、y,forecastingresearchandpracticehasbeendominatedbystatisticalmethods.Morerecently,neuralnetworksandotheradvancedmethodsonpredictionhavebeenusedinfinancialdomains[1-3].Aswegettoknowmoreaboutthedynamicnatureofthefinancialmarkets,theweaknessesoftraditionalmethodsbecomeapparent.Inthelas
6、tfewyears,researchhasfocussedonunderstandingthenatureoffinancialmarketsbeforeapplyingmethodsofforecastingindomainsincludingstockmarkets,financialindices,bonds,currenciesandvaryingtypesofinvestments.Peters[4]notesthatmostfinancialmarketsarenotGaussianinnatureandtendtohavesharperpeaks
7、andfattails,aphenomenonwellknowinpractice.Inthefaceofsuchevidence,anumberoftraditionalmethodsbasedonGaussiannormalityassumptionhavelimitationsmakingaccurateforecasts.OneofthekeyobservationsexplainedbyPeters[4]isthefactthatmostfinancialmarketshaveaverylongmemory;whathappenstodayaffec
8、tsthefutureforever.
