An Intelligent Solution Support System for Spatial Modelling and Decision Support

‘Modelling inside GIS’ has been widely researched over the years. This paper argues that such approach lacks appropriate model reusing and management functions because it is often domain-dependent. In this paper, we propose a ‘modelling outside GIS’ approach to design a model management module that can be incorporated with GIS and other systems to perform flexible model retrieval and development. As a first step, an intelligent solution support system (subsystem) that can assist users to select and construct appropriate models for their particular application domains is developed. Some essential mechanisms and procedures involved in the developing this system are proposed and demonstrated. 1. Demands in spatial decision making Geographical information systems (GIS) has emerged as an important information processing and spatial technology. It has helped to regenerate the interests and applications of complex spatial and dynamic models in problem solving and decision making (Bertuglia et al, 1994) . There are two main approaches in current spatial modelling. One commonly adopted approach is ‘modelling inside GIS’ which attempts to incorporate different models into GIS with different coupling and integrating strategies (Batty and Xie, 1994). Another approach is the adoption of decision support systems (DSS) originally developed for operation research and management science (OR/MS) to spatial decision making domains. This is commonly described as spatial decision support systems (SDSS) which is particular useful for illstructured and unstructured spatial problems (Densham and Amstrong, 1994). Recently some intelligent functions 0-7695-0001-3/99 $1 are built into such system to provide knowledge based support in information processing and decision making (Leung, 1997). However, since these approaches are generally developed within GIS or AI environments, and applications are often domain-dependent, they do not provide flexible mechanisms and capabilities for model sharing, reuse and communications among different applications and systems. Thus considerable modelling programming and model development process are often involved when the problem domains and application systems are changed. Prototypes of object-oriented geographical modelling systems (GMS) have recently been proposed with some functions of modelbase management (Bennett, 1997; Raper and Livingstone, 1995). Their approaches support the development and use of dynamic geographical models, and can help users to construct domain specific geooriented simulation models by using icon-based user interface and autonomous modelling agents (Bennett, 1997). However, these systems do not provide appropriate mechanisms for reusing the established models, and do not offer intelligent assistance to users to retrieve and select appropriate models from a modelbase according to their needs. Another drawback is that they do not utilize existing systems such as GIS which has already developed some advanced features in spatial modelling. Model management systems (MMS) have been proposed in OR/MS research to resolve the above problems. Some interesting model management frameworks have been designed such as database approach, graph-based approach, object-oriented approach and knowledge-based approach (see review by Bharadwaj et al, 1992; Blanning 1993). These approaches provide a challenge to the current development of SDSS although spatial modelling problems are generally much more complex than models in operation research and 0.00 (c) 1999 IEEE 1 Proceedings of the 32nd Hawaii International Conference on System Sciences 1999 Proceedings of the 32nd Hawaii International Conference on System Sciences 1999 management science. In this paper, we propose to develop an integrated solution support system (SSS) based on the extension of existing model management framework that can be applied to assist the end-users to select suitable modelling methods, and to construct more complex models for complex spatial problems. 2. Architecture of a solution support system for spatial decision making As illustrated by Densham and Armstrong (1994), a powerful SDSS should facilitate a number of functions such as data analysis, model-based analysis and powerful visual presentation. In recent years, a variety of spatial information systems has been developed. These provide numerous resources upon which an integrated system infrastructure can be developed. With the support of object-oriented programming and system integration techniques, we proposed an integrated SSS environment in Figure 1 which incorporates essential functions of GIS, 0-7695-0001-3/99 $1 database systems, KBS and model management techniques to support overall modelling and decision making processes. The main components of the system are briefly described below.