Identification of variable parameters of a model for the construction of a forecasting algorithm
Abstract
An approach to identification of the mathematical expectation of acceleration of values change of data samples, which varies according to an unknown law, is presented in this article. An estimation method of mathematical expectation of values acceleration of change of data samples is developed, which is used to construct a forecasting algorithm based on the Kalman filter. An imitation modeling was performed, which showed the effectiveness of the suggested approach. The forecasting algorithm model based on the Kalman filter, autoregressive model and autoregressive moving average model are constructed using the daily average of the lead prices on the London Metal Exchange, and forecasting is done on the same data set. A comparative analysis of presented models, using the characteristics of forecasting values showed the advantage of the forecasting algorithm based on the Kalman filter.References
Astrom K.J., Eykhoff P. System identification – a survey // Automatica. –1971. – 7. – Р. 123–162.
Fil’tratsiya i stokhasticheskoye upravleniye v dinamicheskikh sistemakh: pod red. K.T. Leondesa. – M.: Mir, 1980. – 407 s.
Anderson W.N., Kleindorfer G.B., Kleindorfer P.R., Woodroofe M.B. Consistent Estimates of the Parameters of a Linear System // The Annals of Mathematical Statistics. – 1969. – 40. – Р. 2064–2075.
Zgurovskiy M.Z., Podladchikov V.N. Аnaliticheskiye metody kalmanovskoy fil’tratsii dlya sistem s apriornoy neopredelёnnost’yu. – K.: Naukova dumka, 1995. – 283 s.
Podladchikova T., Van der Linden R. A Kalman Filter Technique for Improving Medium-Term Predictions of the Sunspot Number // Solar Physics. – 2012. – 277. – Р. 397–416.
Bratus' O.V., Podladchikov V.M. Pobudova bahatovymirnoyi modeli na osnovi fil'tra Kalmana y analiz alhorytmiv otsinyuvannya yiyi parametriv // Naukovi visti NTUU "KPI". – 2013. – # 5. – S. 28–34.
Lukashin YU.P. Аdaptivnyye metody kratkosrochnogo prognozirovaniya vremennykh ryadov: ucheb. posobiye. – M.: Finansy i statistika, 2003. – 416 s.
Bidyuk P.I., Korshevnyuk L.O. Proektuvannya komp"yuternykh informatsiynykh system pidtrymky pryynyattya rishen': navch. posibnyk. – K.: NNK "IPSA" NTUU "KPI", 2010. – 340 s.
Bendat Dzh., Pirsol А. Prikladnoy analiz sluchaynykh dannykh / Per. s angl. d-ra fiz.-mat. nauk V.E. Prival’skogo i А.I. Kochubinskogo pod red. Аkad. I.N. Kovalenko. – M.: Mir, 1989. – 540 s.
Ofitsiynyy sayt Londons'koyi birzhi metaliv. Rozdil "Istorychni dani": http://www.lme.com/historical_data.asp.
