Comparison of some ANN based forecasting methods implemented on short time series

ANN versus Grey theory based forecasting methods implemented on short time series

“…These were reported in [1,2,3,4] and may be characterized as follows: use of simple ANNs (only several neurons in the hidden layer, between one and ten), and use of short time series. There are two reasons for the later.…”

“…Second, as it is frequently clamed in the forecasting literature, only the most recent data are of prime importance in the behavior of many natural and social phenomena. In the next, we will briefly describe our ANN solution as given in [1,2,3,4].…”

“…N our recent study [1] we introduced two ANN structures convenient for prediction based on short time series. Namely, the FFAP and the TCR structure were successfully implemented to prediction of quantities of obsolete computers in the USA.…”

“…After publishing these results we implemented the new methods to prediction in two new areas: electricity load [2] and microelectronic developments [3,4]. Among other, in [1,2,3,4] the claim "One major criticism about the NN model is that it demands a great deal of training data and relatively long training period for robust generalization" [5] was denied. Namely, neural networks with small numbers of hidden neurons were implemented and prediction were produced based on extremely reduced sets of data within very short training time periods.…”

“…While comparisons were given in [1,2,3,4] with some other forecasting methods no discussion of the Grey theory [6,7] and, accordingly, no adequate comparisons, were presented.…”