Speaker: Professor Robert Fildes, Dept of Management Science,Lancaster university
Title: Business Forecasting and statistical modelling: optimally combining disparate information <http://www.maths.lancs.ac.uk/department/events/rss/talk.2008-10-16.8467244406>
Time: Nov 27, 2008, Thursday , 5pm (Tea/cookies at 4:45pm)
Venue: Fylde SCR, Lancaster University
All welcome!
###########
The most common approach to forecasting demand in support of supply
chain planning uses a simple statistical model to produce a forecast.
This is subsequently judgmentally adjusted by the company's demand
planners to take into account any exceptional circumstances expected
over the planning horizon using a forecasting support system (FSS).
Based on company case studies of more than 5000 adjusted forecasts
this presentation examines the improvements that can be achieved over
the statistical system forecast by effective adjustment. Overall,
adjustment improves accuracy. The paper compares company forecasts, the
unadjusted data from the companies' established FSSs and more advanced
statistical models. The company forecasts all suffer from similar
errors, depending on the direction and strength of market information
that the companies believe they posses.
The company forecasts can be shown to be inefficient and
unnecessarily inaccurate. Various combining algorithms can be used to
improve forecast accuracy. In practice it will prove difficult to
capitalize on these inefficiencies. A possible route forward is through
improved forecasting software design, rather than improved statistical
modelling.
#############
|