Top Banner
CÁC PHƯƠNG PHÁP DỰ BÁO CƠ BẢN
62

Chuong 2 - Cac Phuong Phap Du Bao Co Ban

Nov 05, 2015

Download

Documents

do do

các phương pháp dự báo cơ bản
Welcome message from author
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
  • CC PHNG PHP D BO C BN

  • Ni dungM t d liu bng thBin tr ca chui thi giano lng chnh xc ca d boChuyn dng bin s v iu chnh d liu

  • M T D LIU BNG TH

  • Chui thi gian v d liu choD liu chui thi gian: mt dy cc quan st theo thi gian D liu cho: tt c cc quan st trong cng mt thi im

  • Bng 1: sn lng bia ca c theo thng t thng 1 nm 1991 n thng 8 nm 1995

  • Bng 2: nc sn xut, mc tiu thu nhin liu v gi c mt s loi xe t

  • th thi gian (time plot)M t s thay i ca d liu theo thi gian

  • Hnh mu d liuMu d liu n nh (Horizontal pattern) Mu d liu thi v (Seasonal pattern)Mu d liu chu k (Cyclical pattern) Mu d liu xu th (Trend pattern) Co 4 hnh mu thay i ca d liu theo thi gian

  • D liu n nhD liu xu thD liu ma vD liu chu k

  • Hnh mu d liuMu d liu n nh: d liu dao ng quanh mt gi tr trung bnh c nhMu d liu thi v: d liu b nh hng bi nhng yu t ma v (qu, thng, tun)

  • Hnh mu d liuMu d liu xu th: d liu tng hoc gim trong mt khong thi gian diMu d liu chu k: d liu dao ng quanh ng xu th tng hay gim khng theo mt khong thi gian c nh no

  • Kt hp cc mu d liuTrong thc t, mt chui thi gian thng kt hp nhiu mu d liu

  • S thi v (seasonal plot)Ct cc chui thi gian theo tng giai on thng lp li v biu din chng trn cng mt trc thi gianGip pht hin yu t thi v

  • S phn tn (scatter plot)Biu din mi quan h ph thuc gia hai bin sGip pht hin mi quan h, mc quan h, chiu quan h (thun, nghch)

  • Bin tr ca chui thi gian

  • Bin tr ca chui thi gianTrong mt chui thi gian, quan st Yt-k c gi l tr k giai thi k ca quan st YtBin s cha cc quan st tr Yt-k c gi l mt bin tr ca bin s cha cc quan st Yt

  • T hip phng sai v t tng quanT hip phng sai v t tng quan l s m rng ca hip phng sai v tng quan vo chui thi gian nghin cu s ph thuc ca mt bin s vi cc bin tr ca n

  • Hip phng sai v tng quanHip phng saiH s tng quan

  • T hip phng sai v t tng quanT hip phng saiT tng quan

  • S lng VCR bn trong nm 2000 ti cc ca hiu hng in t

    Time tMonthOriginal dataYtY lagged one period Yt-1Y lagged two period Yt-2123456789101112JanuaryFebruaryMarchAprilMayJuneJulyAugustSeptemberOctoberNovemberDecember123130125138145142141146147157150160123130125138145142141146147157150160

    123130125138145142141146147157150160

  • Hm t tng quan (ACF)Hm ACF cho bit h s t tng quan ti cc mc tr khc nhauACF c biu din bng s t tng quan (correlogram)

  • Sn lng bia c

  • O LNG CHNH XC CA D BO

  • Ct 2: Cc quan st trong 8 thng (t1- t8 nm 1995)Ct 3: Cc gi tr d bo, tnh bng cch ly trung bnh ca cng thng trong 4 nm.

    VD: gi tr d bo cho thng 1 nm 1995 l trung bnh ca thng 1 cc nm 1991, 1992, 1993, 1994.Sn lng bia hng thng ca c

    T YtFt1138150.252136139.53152157.254127134.551511386130127.57119138.258153141.59140.510167.25

  • o lng sai sYt l quan st thc trong k t v Ft l gi tr d bo cho cng kSai s c nh ngha: et =Yt-Ft

  • Cc o lng sai s c bn MAE = MSE =

  • tY FY-F|Y-F |(Y-F)212345678138136152127151130119153150.25139.5157.25134.5138127.5138.25141.5-12.25-3.5-5.25-7.5132.5-19.2511.512.253.55.257.5132.519.2511.5150.062512.2527.562556.251696.25370.5625132.25Tng -29.7583.751140.20ME = -29.75/8 = -3.72MAE = 83.75/8 = 10.47MSE = 1140.20 /8 = 142.52

  • Sai s tng iSai s phn trm

    Sai s phn trm trung bnh

    Sai s phn trm tuyt i

  • MPE = -26.0/8 = -3.3%MAPE = 62.1/8 = 7.8%62.1-26.0Tng8.92.63.5138.61916.27.5-8.9-2.6-3.5-138.61.9-16.27.512.253.55.2516.5132.519.2511.5-12.25-3.5-5.25-16.5132.5-19.2511.5150.25139.5157.25143.5138127.5138.25141.513813615212715113011915312345678|Yt-Ft |Yt-Ft FtYtT

  • So snh cc phng php d boLy mt phng php n gin lm c s so snh kt qu d bo vi nhng phng php phc tp hn.Phng php d bo Nave 1 hay NF1: Gi tr k sau bng gi tr k ngay trc : Ft+1 = Yt

  • Tnh sai s phn trm ca phng php NF1

    TYtFt (NF1)|Yt - Ft| 1234567813813615212715113011915318213813615212715113011944216252421113431.91.510.519.715.916.29.222.2Tng117127.1MAE = 117/8 = 22.1 MAPE = 127.1/8 = 15.9%

  • H s Theil UH s U c pht trin bi Theil (1966) cho php so snh so snh kt qu ca cc phng php d bo vi phng php NF1

  • Thay i tng i ca gi tr d bo

    Thay i tng i ca gi tr thc t

  • MAPE ca phng php d bo cn so snhMAPE ca phng php d bo NF1

  • Kt lunU=1: hai phng php nh nhauU1: phng php dang s dng khng tt hn NF1.

  • H s Theil U

  • Hm sai ACF ca s d boHm t tng quan ca sai s d bo cho bit c cn thay i mang tnh h thng trong cc sai s hay khng.

  • CHUYN DNG BiN S V iI CHNH D LiU

  • Ba dng iu chnh Chuyn dng bin s bng cc hm ton iu chnh d liu do nh hng ca lch thi gian iu chnh tc ng ca lm pht hoc mc tng dn s

  • Chuyn dng bin sDng cc hng ton hc chuyn dng bin s. Mc ch l lm n nh mc bin thin ca gi tr trong bin s.Quy trnh:Chuyn dang bin sD bo da trn d liu chuyn dngChuyn dng ngc li d bo bin s ban u

  • Mt s dng chuyn i thng dngCn bc 2Cn bc 3Hm logm nghch oChuyn dngChuyn ngc

  • Dng thc chungChuyn dng hm mChuyn dng ngc

  • D liu gc

    Yt

    Ln(Yt)

    Jan-561254Feb-561290Mar-561379Apr-561346May-561535Jun-561555Jul-561655Aug-561651Sep-561500Oct-561538

    Jan-56125435.411862427.1341Feb-56129035.916569997.1624Mar-56137937.134889267.2291Apr-56134636.687872667.2049May-56153539.179076057.3363Jun-56155539.433488317.3492Jul-56165540.681691217.4116Aug-56165140.632499317.4091Sep-56150038.729833467.3132Oct-56153839.21734317.3382

  • Variation increases

  • Cn bc 2

  • Log

  • iu chnh do lch thi gianThay i trong chui thi gian c th do s thay i v s ngy theo lch hoc s ngy lm vic trong thng.

  • iu chnh thay i ngy trong thngKhc bit gia s ngy ca thng di nht v ngn nht l (31- 28)/30 = 10%.

  • S ngy lm vic trong thngS ngy lm vic khc nhau trong cc thng

  • Sn lng s hng thng (365.25/12)= ------------ x 589 31

  • Sn lng sa

  • Sn lng sa iu chnh

  • iu chnh do lm pht hay pht trin dn sV d 1: $1000 nm 2007 khc vi $1000 nm 1997. iu tit ta c th quy v gi tr ti mt mc thi gian.V d 2:D bo s ngi s dng phng tin cng cng qua cc nm phi tnh n mc tng dn s.Thay v d bo tng s ngi s dng, ta d bo t l ngi s dng trn tng s dn.

    Data patterns, including components such as trend, seasonality and irregularity, can be studies using the autocorrelation analysis.