A new approach to the spatial analysis of temporal change using nodes and neural nets

The term 'temporal' in spatial analysis has a number of potential meanings, each of which requires an alternative approach for the provision of analytic support. Much of the present work in spatio-temporal information is concerned with transaction versioning. Object based representations often demand a high level of initial understanding of object relationships. Many GIS users are seeking to understand the object relationships over time, past, present and future; their research focus is how real-world features interact in time and space. Despite this requirement, little present work will support this requirement to understand the drivers of change, rather than simply to report what changed. A number of workers have / are attempting to formalise a theory of spatio-temporal reasoning (eg Hermosilla, 1994, Qian, et al, 1997, Claramunt et al, 1997), in the most part working from a theoretical abstraction. Worboys, 1998, uses a problem oriented approach, as does Halls and Miller (1995, 1996). Representation of change over time by means of spline curves offers possibilities for this type of work, neural networks are explored as an implementation solution. We show that the AURA neural network architecture offers particular hope and that the proposals of Yeh & de Cambray and Halls & Miller need to be recast in terms of the AURA architecture.