Prediction with Neural Networks

What is Neural Network Prediction?

Neural network prediction uses trained artificial neural networks to forecast future values from historical data. A multi-layer feed-forward network learns patterns in a time series through backpropagation and then applies that learned model to predict upcoming values — for example, stock prices or currency exchange rates.

This tutorial introduces prediction using artificial neural networks. In particular, it describes time-series prediction using multi-layer feed-forward neural networks.

First, prediction itself is introduced together with a basic classification of prediction types. After that, prediction using neural networks (NNs) is explained. The main focus is on creating a training set from a time series. To illustrate the topic, interactive demos are available that show how a training set is created and how prediction works with a backpropagation neural network, both for analytic functions and for financial data.

To continue, press the Next button, or skip the introductory text and directly try the prediction online.

New! In March 2026, the interactive demonstrations were rewritten for modern web, so you can try them again now.

References

Werbos, P. (1974). Beyond Regression: New Tools for Prediction and Analysis in the Behavioral Sciences. PhD thesis, Harvard University.
Rumelhart, D. E., Hinton, G. E., & Williams, R. J. (1986). Learning representations by back-propagating errors. Nature, 323, 533–536. DOI: 10.1038/323533a0