This 2001 report by Rod Katz offers an overview of various methods used to forecast demand for bicycle infrastructure outlining the relative merits of each.

There is a variety of demand modelling techniques at the disposal of transport planners and bicycle planners in particular. Many of these are dependent on obtaining appropriate data and having the technological expertise available to implement the methods.
Some of the easier methods such as comparison studies can be used to support proposals at the local level. For more widespread bicycle facilities and programs involving significant expenditure, a more rigorous forecasting framework is likely to be required.
Such a framework is likely to need to be developed alongside regional travel models. These models are not currently equipped to evaluate the case for improvements in the bicycle network because of their reliance on calibration based on existing travel patterns – bicycle use is generally too insignificant in Australian cities to provide the basis for calibration within regional travel models.
In order to develop tractable models of the role of bicycles within existing and forecast transport systems it would be possible to develop a number of the techniques described in this guide. In particular, the use of sketch planning techniques with simplified networks and using behavioural parameters from discrete choice models could be a profitable area for regional planners. These could be used to alongside some of the visioning exercises that are becoming increasingly popular (eg. Queensland Transport and Main Roads (1999)).
The availability of GIS modelling techniques also offers considerable scope for developing a better understanding of where bicycle demand can be expected and relating that demand to the urban landscape in a readily comprehensible fashion.
GIS systems can be used to enhance local area bicycle plans and local area integrated transport planning.