Applied Spatial Interaction Modelling

The aim of this chapter is to reflect on progress with spatial interaction modelling in a wide range of commercial environments and to focus on how applied work has fed back some important challenges to make spatial interaction models work more effectively. Such applied models build on thirty years of academic model development beginning with Wilson’s derivation of the model through entropy maximisation (Wilson 1967, 1970). We briefly review theoretical progress with this model in section 8.2. In section 8.3 we turn our attention to applied modelling and discuss why such models have been so successful in a commercial environment and how they are able to address a wide variety of client requirements. In section 8.4 we discuss the increasing availability of better quality data which enables calibration to be undertaken more effectively. In turn, the availability of better data can shape the form of the models themselves (through greater disaggregation), although it quickly becomes apparent when dealing with applied problems that models are ‘context’ dependent, usually requiring fine tuning to work in specific sectors. We thus also explore the theoretical extensions to the models based on commercial applications and the data they bring. A particular interest of ours, and essential in a commercial context, is the ability to demonstrate that spatial interaction models ‘work’ . This is a subject that has been largely ignored by the academic community except in the narrow technical sense of calibration. We shall discuss this issue in section 8.5. In section 8.6 we go through a typical application of the models, showing how a user would interface with the information system to analyse the impacts of the changes they simulate on the computer. Concluding comments are given in section 8.7.