Neural networks in time series forecasting

Rohit Dhawan

This research deals with the issues of why data has to be deseasonalised first before applying it to an artificial neural network (ANN) model for the use of forecasting. Since neural networks are expected to capture non-linear behaviour such as seasonality, they still do not do as well as when they get preprocessed data. The work deals with exploring reasons as to why this happens and finding new ANN models and techniques by which ANNs could incorporate seasonality.

Supervisor/s

Professor Marcus O'Connor

28176