Unrivalled Accuracy in Forecasting weekly and daily time series with Neural Networks
Published 17 February 2009
FORECASTING SUCCESS STORY: Unrivalled Accuracy in Forecasting weekly and daily time series with Neural Networks
ESTSP’08 & NN5 forecasting competition; Projects in Call Centre Forecasting
The Forecasting Centre has shown preeminent forecasting accuracy on so called ‘high-frequency’ time series of weekly and daily data. Instead of aggregating time series into weeks or months and using rigid day-of-the-week patterns, the neural network ensembles are capable of learning the structure of the daily time series directly, including effects from bank-holidays, marketing and mailing activities, one-time externals shocks and even weather information! With a proven track-record in electrical load forecasting, this promises new magnitudes of accuracy in daily call centre and retail forecasting, and weekly logistics forecasting.
For additional information please contact Dr. Sven F. Crone, deputy director of the Lancaster Centre for Forecasting.
