Abstract: Time series prognosis of economical indexes is one of the main problems of econometrics. In the paper we study possibility to give an interval prognosis of time series using the set of the best prognostic models. Speaking ‘model’ we mean a combined model of regression and auto-regression. Speaking ‘the best models’ we mean the ordered series of models constructed by the well-known Group Method of Data Handling (GMDH). The proposed simple approach consists in the following: a) one generates the fixed numbers of models on the basis of experimental data b) these models give correspondent prognoses c) the real value is supposed to belong to min-max interval the models provide. We shortly describe the software tool GMDH-Shell (GS) that implements GMDH and the results of experiments with GS. The experimental data are time series of the Gross Domestic Products (GDP) of 100 countries given on the period 1980-2000.

Keywords: GMDH, GMDH Shell, time series prognosis, gross domestic product

ACM Classification Keywords: I.2 Artificial Intelligence

Link:

TIME SERIES PROGNOSIS OF GDP WITH THE SYSTEM GMDH-SHELL (EXPERIMENTAL WORK)

Victor Lebedev

http://foibg.com/ibs_isc/ibs-27/ibs-27-p06.pdf