Group Method of Data Handling (GMDH)-type neural network algorithms are the heuristicself organization method for the modelling of complex systems. GMDH algorithms are utilizedfor a variety of purposes, examples include identification of physical laws, the extrapolation ofphysical fields, pattern recognition, clustering, the approximation of multidimensional processes,forecasting without models, etc. In this study, the R package GMDH is presented to make short termforecasting through GMDH-type neural network algorithms. The GMDH package has options to usedifferent transfer functions (sigmoid, radial basis, polynomial, and tangent functions) simultaneouslyor separately. Data on cancer death rate of Pennsylvania from 1930 to 2000 are used to illustrate thefeatures of the GMDH package. The results based on ARIMA models and exponential smoothingmethods are included for comparison.
展开▼
