*** APOLOGIES FOR CROSS-POSTING ***
Hi
I've published on traffic forecasting accuracy in the past, and I'm
currently researching for a new, updated piece.
Most studies of predictive accuracy have compared point-forecasts with
outturn figures. This is a very harsh (unreasonable?) test. Perhaps we
should establish what is 'reasonable' in terms of forecasting capability?
This is a line of enquiry that I'm exploring.
I have a little spreadsheet (which is very quick to complete) in which I
present two scenarios: one for an existing ('brownfield') road and the other
for a new ('greenfield') road. What level of accuracy do you think is
reasonable for, say, a one-year traffic forecast? Or a five-year one?
The UTSG list will not accept attachments so contact me directly on
[log in to unmask] and I'll send you the spreadsheet. You simply insert your
responses in the appropriate yellow boxes. Eg. +/- 10%, or +/- 50%. It's
very quick and I'd really appreciate your views.
I've given two options in the spreadsheet. One for a toll road (in green)
and the other for a non-tolled ('regular') road (in blue). You may - or may
not - feel that tolling per se changes the level of accuracy which
might/should be expected.
I'll compile my results and will share them with List Subscribers when
complete.
If anyone knows of any similar research, related guidelines etc., please
draw them to my attention.
Many thanks,
Rob Bain
Visiting Research Fellow, University of Leeds
www.robbain.com
|