A new computer oriented method for forecasting land use in an urban area is presented in this study. The street system is described as an arc-node network and extrapolation from past tendencies is performed with the aid of a probabilistic mapping of accessibility. It is thereby possible to consider more than one attraction activity, and to describe the size and shape of an activity rather than dealing with it as a single point. The street system is treated as the primary parameter in the model with the ability to change the network and the location of attraction activities.