Some Results of the Synthesis of GMDH and Factor Analysis for Inductive Modelling. Yuriy V. Dzyadyk. IWIM, Prague, 2007.
Article (in pdf)
Abstract. It is known that GMDH generally requires building and comparison of 2k linear models and choice of the best among them (above k . 2l is a number of all linear arguments selected among basic functions 1, t_(i), t_it_j , : : : , pt_i, e etc, i; j; = 1::l). Since arguments are correlated, a little alteration of input data often results in a models with absolutely different suites of arguments. We propose two steps for application of factor analysis to GMDH. The first step simply consists in using GMDH in the orthogonal basis of factors. On the second, heuristic step we preliminarily obliterate so called unstable and inessential factors. Some versions of this method are realized by means of Java. It was successfully used for modelling and forecasting of extremely unstable molybdenum prices: monthly prices in 2004–05, and annual prices in 1975–98.
Keywords. inductive modelling, IWIM 2007, factor analysis, price forecasting, market forecasting, MAS forecasting
References.
- Ivakhnenko A.G., Ivakhnenko G.A.: The Review of Problems Solvable by Algorithms of the Group Method of Data Handling (pdf)
- Koppa Yu.V., Stepashko V.S.: A Comparison of the Forecasting Properties of Regression Types Models versus GMDH (pdf) (in Russian)
- MEPS (International) Ltd. – Independent Steel Industry Analysts, Consultants, Steel Prices, Reports and Publications. Prices.htm World Stainless Steel Product Prices
- Energy Information Administration. Official Energy Statistics from the U.S. Government. World Nominal Oil Price Chronology: 1970-2005.
- Tymashova L.A., Dzyadyk Yu.V., Leshchenko V.A., Bondar L.A.: An Intelligent System for Price Forecasting (in Ukrainian) // Problemy vprovadzhennya informacijnyh tehnologij v ekonomici. Tezy dopovidej VI mizhnarodnoyi naukovo-praktychnoyi konferenciyi. Irpin’, 2007
