Time Series Forecasting

《Time Series Forecasting》由会员上传分享，免费在线阅读，更多相关内容在学术论文-天天文库。

1、TimeSeriesForecastingMatiasG.EnzTimeSeriesForecastingQuickreviewofforecastingDifferentApproachestoForecastingIntroductiontotimeseriesMeasuringForecastErrorDescribingatimeseriesForecastingmethodsforleveldemandWhatisForecasting?Forecastingisthepredi

2、ction,projectionorestimationoftheoccurrencesofuncertainfutureeventsorlevelsofactivity.Examples:WeatherthisweekendStockpricesNextmonth’ssalesinregionX…ResultsofpoorforecastingFundamentalForecastingApproachesWhatdoesthefuturehold?QualitativeRelatio

3、nalTimeSeriesTimeSeriesApproachesConsistsofdatathatarecollected,recorded,orobservedoversuccessiveincrementsoftimeAssumespastbehaviorisgoingtoberepeatedinthefutureSearchforpatternsinthedata--decompositionintodifferentcomponentsManytimeseriestechnique

4、sDescribingaTimeSeriesWecanthinkofahistoricaltimeseriesasbeingcomposedof“pattern”and“noise.”the“pattern”canincludestationarity,trend,seasonality,oracycle,the“noise”israndomwithzeromeanandseriallyuncorrelatedThegoalofforecastingistoforecastthepattern.

5、Noiseisrandomandcan’tbeforecasted.ComponentsoftheTimeSeriesOriginaltimeseriesDecomposedtimeseriesCyclecycleSeasonalseasonalityDemandDemandTrendtrendLevellevelTimeNoisenoiseTimeDemandcanbeexpressedasafunctionofthesecomponents,forexample,D(t)=(L+Tt)[S(t)]

6、+rChartstakenfrom“OperationsManagement,”R.G.Schroeder,IrwinMcGraw-Hill,2000ForecastOptimalityAforecastisoptimalif:allpatternhasbeendiscoveredonlyerrorcomesfromthenoisetermForecastfitreferstohowtheforecastmatches-upagainstpastdataover-fitting(forecas

7、tingtherandomness)isbadForecastErrorAmeaningfulforecastconsistsof2values:Pointestimate+errorForecasterrorinformationisusedfor:SelectingforecastingmethodsEstimatingsafetystocksImprovingforecastingprocessEvaluatingriskSomeTimeSeriesTechniquesSimple

8、ComplexMovingaveragesSpectralanalysisExponentialBox-Jenkinssmoothing(ARIMA)DecompositionKalmanFiltersFourierAnalysis?EvaluatetheforecastusingnewPastdataisavailable.demanddataasitcomesin....DDDDDDD...t-3t-2t-1tt+1t+