Informatica e diritto, XIX Annata, Vol. II, 1993, n. 2, pp. 53-73

Klaus Köhler

Forecasting Economic Time-Series with Neural Networks

Prevedere time-series economiche con le reti neurali
Relazione al Seminario "The Computer and Vagueness: Fuzzy Logic and Neural Nets"
(Computer e vaghezza: logica fuzzy e reti neurali), Monaco di Baviera 20 novembre 1992

Sommario: 1. Introduction. 2. General Properties of Neural Networks. 3. Pattern Association in Economic Time-Series Forecasting. 4. Using Backpropagation Networks for Pattern Association. 4.1. Pattern Association with Feedforward Networks. 4.2. Learning in Backpropagation Networks. 5. Difficulties in Forecasting Time Semes. 5.1. General Forecasting Problems. 5.2. The Generalization Problem. 6. How to Find a Backpropagation Net with Optimal Generalization Performance. 6.1. Selecting input and output data. 6.2. Preprocessing and Postprocessing. 6.3. Network Topology. 6.4. Pruning. 6.5. Stopped Training. 6.6. Penalty Terms. 7. Results. 8. Conclusions. References.