Analisi delle serie temporali

Basicknow

Atimeseriesisasequenceofnumbersinchronologicalorder.

Characteresoftimeseries:

1.Arealisticandtruesetofdata,notobtainedthroughexperimentsinmathematicalstatistics.Sinceitistrue,itisastatisticalindicatorthatreflectsacertainphenomenon.Therefore,behindthetimeseriesisthelawofchangeofacertainphenomenon.

2.Dynamicdata.

Thebasicstepsoftimeseriesmodelingare:

1.Obtainthetimeseriesdynamicdataoftheobservedsystembymethodssuchasobservation,survey,statistics,andsampling.

2.Drawcorrelationgraphsbasedondynamicdata,conductcorrelationanalysis,andfindautocorrelationfunction.Thecorrelationdiagramcanshowthetrendandcycleofchanges,andcanfindjumppointsandinflectionpoints.Jumppointsareobservationsthatareinconsistentwithotherdata.Ifthejumppointisthecorrectobservationvalue,itshouldbetakenintoaccountwhenmodeling,ifitisanabnormalphenomenon,thejumppointshouldbeadjustedtotheexpectedvalue.Theinflectionpointisthepointatwhichthetimeseriessuddenlychangesfromanupwardtrendtoadownwardtrend.Ifthereisaninflectionpoint,differentmodelsmustbeusedtofitthetimeseriessegmentallyduringmodeling,suchasathresholdregressionmodel.

3.Identifyasuitablerandommodelandperformcurvefitting,thatis,useageneralrandommodeltofittheobservationdataofthetimeseries.Forshortorsimpletimeseries,trendmodelsandseasonalmodelspluserrorscanbeusedforfitting.Forstationarytimeseries,generalARMAmodel(autoregressivemovingaveragemodel)anditsspecialcaseautoregressivemodel,movingaveragemodelorcombined-ARMAmodelcanbeusedforfitting.Whentherearemorethan50observations,theARMAmodelisgenerallyused.Fornon-stationarytimeseries,theobservedtimeseriesmustbefirstdifferentiatedintoastationarytimeseries,andthenanappropriatemodelisusedtofitthedifferenceseries.

Characteres

Timeseriesanalysisisoneofthequantitativeforecastingmethods.Itincludesgeneralstatisticalanalysis(suchasautocorrelationanalysis,spectrumanalysis,etc.),theestablishmentandinferenceofstatisticalmodels,andtheoptimalprediction,controlandfilteringoftimeseries.Classicalstatisticalanalysisassumestheindependenceofdataseries,whiletimeseriesanalysisfocusesontheinterdependenceofdataseries.Thelatterisactuallyastatisticalanalysisoftherandomprocessofdiscreteindicators,soitcanberegardedasacomponentofrandomprocessstatistics.Forexample,therainfallofthefirstmonth,thesecondmonth,...,theNthmonthinacertainareaisrecorded,andtherainfallinthefuturemonthscanbeforecastedbyusingthetimeseriesanalysismethod.

Basicidea:Basedonthesystem'slimited-lengthoperatingrecords(observationdata),establishamathematicalmodelthatcanmoreaccuratelyreflectthedynamicdependenciescontainedinthesequence,anduseittopredictthefutureofthesystem.

Basicprinciples:Oneistorecognizethecontinuityofthedevelopmentofthings.Usingpastdata,wecaninferthedevelopmenttrendofthings.Thesecondistoconsidertherandomnessofthedevelopmentofthings.Thedevelopmentofanythingmaybeaffectedbyaccidentalfactors.Forthisreason,theweightedaveragemethodinstatisticalanalysisshouldbeusedtoprocesshistoricaldata.

Features: Simplex et facilis, facilis ad dominum, sed pauper cura, plerumque solum idoneum ad breve tempus praemittit.

Classification

Accordingtoitscharacteristics,thetimeserieshasthefollowingmanifestations,andproducescorrespondinganalysismethods:

1.Long-termtrendchanges:Affectedbyacertainbasicfactor,thedatashowsacertaintendencywhenitchangesovertime,anditsteadilyincreasesordecreasesaccordingtoacertainrule.Theanalysismethodsusedare:movingaveragemethod,exponentialsmoothingmethod,modelfittingmethod,etc.

2.Seasonalcyclechanges:Affectedbyfactorssuchasseasonalchanges,thesequencechangesregularlyaccordingtoafixedcycle,alsoknownasthebusinesscycle.Methodused:seasonalindex.

3.Cyclicchanges:fluctuantes commutationes with irregularcycles.

4.Randomchanges: Sequentiae mutationum causarum incertorum causa est.

Timeseriesanalysismainlyincludesdeterministicchangeanalysisandrandomchangeanalysis.Amongthem,thedeterministicchangeanalysisincludestrendchangeanalysis,cyclechangeanalysis,andcyclechangeanalysis.Randomchangeanalysis:AR,MA,ARMAmodels,etc.

Specificmethods

Deterministictimeseriesanalysis

Thepurposeofdeterministictimeseriesanalysis:toovercometheinfluenceofotherfactors,simplymeasureacertaindeterministicfactoronthesequenceTheinfluenceofvariousdeterministicfactorsandtheircomprehensiveinfluenceonthesequenceareinferred.

Thepurposeoftimeseriestrendanalysis:Sometimeserieshaveverysignificanttrends.Thepurposeofouranalysisistofindthistrendinthesequenceandusethistrendtomakereasonablepredictionsforthedevelopmentofthesequence.

Commonmethods:trendfittingmethodandsmoothingmethod.

Thetrendfittingmethodistousetimeastheindependentvariableandthecorrespondingsequenceobservationvalueasthedependentvariabletoestablisharegressionmodelofthesequencevaluechangingwithtime.Includinglinearfittingandnonlinearfitting.

Theuseoccasionoflinearfittingistheoccasionwherethelong-termtrendshowslinearcharacteristics.Theparameterestimationmethodisleastsquareestimation.

Themodelis, ,.

Theuseoccasionsofnonlinearfittingareoccasionswherethelong-termtrendshowsnon-linearcharacteristics.Theideaof​​parameterestimationistoconverteverythingthatcanbeconvertedintoalinearmodelintoalinearmodel,andusethelinearleastsquaremethodtoestimatetheparameters.Ifitcan'tbeconvertedtolinear,useiterativemethodtoestimatetheparameters.

Themodelsinclude,,,etc.

Smoothingmethodisacommonlyusedmethodfortrendanalysisandforecasting.Itusessmoothingtechnologytoweakentheinfluenceofshort-termrandomfluctuationsonthesequenceandsmooththesequence,therebyshowingthelawoflong-termtrendchanges.

Timeseriesforecastingmethod

Timeseriesforecastingmethodcanbeusedforshort-termforecasting,mid-termforecastingandlong-termforecasting.Accordingtothedifferentmethodsofdataanalysis,itcanbefurtherdividedinto:simplesequentialtimeaveragemethodandweightedsequentialtimeaveragemethod.

Simpleaveragemethod:alsoknownasarithmeticaveragemethod.Thatis,thestatisticalvalues​​ofanumberofhistoricalperiodsaretakenastheobservedvalues,andthearithmeticaverageiscalculatedasthepredictedvalueforthenextperiod.Thismethodisbasedonthefollowinghypothesis:"Itwasthesameinthepast,anditwillbethesameinthefuture."Itequatesandaveragesshort-termandlong-termdata,soitcanonlybeappliedtotrendforecastswherethingshavenotchangedmuch.Ifthingsshowacertainupwardordownwardtrend,thismethodshouldnotbeused.

Weightedaveragemethod:weightthehistoricaldataofeachperiodaccordingtothedegreeofshort-termandlong-terminfluence,andcalculatetheaveragevalueasthenextforecastvalue.

Randomchangeanalysis

Therandomtimeseriesmodel(timeseriesmodeling)referstoamodelbuiltusingonlyitspastvalues​​andrandomdisturbanceterms,anditsgeneralformis

.Takethelinearequation,theone-periodlag,andthewhitenoiserandomdisturbanceterm().

Themodel will bea first-orderautoregressiveprocessAR(1).

Thegeneralp-orderautoregressiveprocessAR(p) est.

Iftherandomdisturbancetermisa albumine (), tunc formula (1) dicitur apureAR(p) processus (purus (p) processus), denotatur.

Ifisnota whitenoise, plerumque consideratur tobeaq-ordermovingaverageprocessMA(q):.

CombinepureAR(p)withpureMA(q)togetageneralautoregressivemovingaverage(aunoregressivemovingaverage)processARMA(p,q).

Theformulashows:

1.Arandomtimeseriescanbegeneratedbyanautoregressivemovingaverageprocess,thatis,theseriescanbegeneratedbyitsownpastorlagvalueandrandomdisturbancetermsToexplain.

2.Ifthesequenceisstationary,thatis,itsbehaviordoesnotchangeovertime,thenwecanpredictthefuturethroughthepastbehaviorofthesequence.Thisisexactlytheadvantageoftherandomtimeseriesanalysismodel.ItshouldbenotedthatnoneoftheaboveARMA(p,q)modelscontainsaconstantterm.Ifaconstanttermisincluded,theconstanttermdoesnotaffecttheoriginalpropertiesofthemodel,becausethemodelcontainingtheconstanttermisconvertedtothemodelwithouttheconstanttermthroughappropriatedeformation.

Mainuses

Timeseriesanalysisiscommonlyusedinthemacro-controlofthenationaleconomy,regionalcomprehensivedevelopmentplanning,businessmanagement,marketpotentialprediction,meteorologicalforecast,hydrologicalforecast,earthquakeprecursorforecast,Croppestsanddiseasesforecast,environmentalpollutioncontrol,ecologicalbalance,astronomyandoceanography.Itmainlyincludesresearchandanalysisfromthefollowingaspects.

Systemdescription

Accordingtothetimeseriesdataobtainedfromtheobservationofthesystem,thecurvefittingmethodisusedtoobjectivelydescribethesystem.

Systemanalysis

Whentheobservationsaretakenfrommorethantwovariables,thechangesinonetimeseriescanbeusedtoexplainthechangesintheothertimeseries,Soastogaininsightintothemechanismofagiventimeseries.

Predictthefuture

Generally,theARMAmodelisusedtofitthetimeseriestopredictthefuturevalueofthetimeseries.

Decisionandcontrol

Accordingtothetimeseriesmodel,theinputvariablescanbeadjustedtokeepthesystemdevelopmentprocessatthetargetvalue,thatis,whentheprocessispredictedtodeviatefromthetargetThenecessarycontrolcanbecarriedout.

Related Articles
TOP