Feed-Forward Neural Networks and Their Applications in Forecasting

Neural networks are computational models with the capacity to learn, to generalize, or to organize data based on parallel processing. Among all kinds of networks, the most widely used are multi-layer feed-forward neural networks that are capable of representing non-linear functional mappings between inputs and outputs and are hailed as  “Universal Approximators”. These networks can be trained with a powerful and computationally efficient method called error back-propagation.

In this thesis, a multi-layer feed-forward neural network based gas load forecast model - the TellFuture load forecast system, is built with Java to show how neural networks work in forecasting. It is known that gas load depends on many factors such as weather, calendar, and other economic information. The model will capture those effects, reflect them within the system, and provide valuable future forecasting data. Similar models can be built to solve problems in other fields as long as the correct relationship between the inputs and the outputs can be captured.