Intelligent Decision Support System – An Integration

In geographical analysis, the demand on domain specific knowledge, technical know hows, and accessibility to data is tremendous. It is difficult to make good decisions without powerful systems to provide, in an integrative way, supports in various phases of the decision making process. Taking the three phase process as an example, the system needs to support, for instance, problem diagnosis, access and scanning of databases, interpretation and monitoring of information in the intelligence phase; generation of alternatives and prediction in the design phase; and analysis of scenarios (e.g., what ifs), explanation, and justification in the choice phase. It is almost useless if such a decision support system has no intelligence to handle information efficiently and to apply the right kind of knowledge to assist the decision making process in a user friendly manner.

For a system to be able to support decision making, it has to possess a certain level of intelligence. Therefore, an SDSS should be able to reason with structured and loosely structured knowledge. They should be able to manage data and user communication efficiently and effectively. Their development calls for the utilization of artificial intelligence and knowledge engineering methods to represent and infer with spatial knowledge; software engineering techniques to manage systems development, information and control flows of models and data; and spatial information system technologies to process and display data. All these have to be integrated in a seamless manner.

To be generic and economical, instead of building an SDSS for a specific problem from scratch, we need to develop an SDSS development environment (shell or generator) so that experts can use it to build effectively and efficiently a variety of domain specific SDSSs. That is, we should have a general development tool which decision makers can use to customize, modify, adapt, and evolve SDSS for solving specific spatial problems.

The general architecture of such SDSS shell is depicted in Figure 2. The core of the system is an expert system shell, which, standing alone, can be used as an expert system development tool. When used in an SDSS, the expert system shell directs control flows and information flows. It provides facilities to represent and store domain specific knowledge acquired from experts or learning examples. It can also contain meta knowledge for inference control, systems and user interface, and external communication. The shell has in its possession inference mechanisms for reasoning with loosely structured spatial knowledge. It is the brain of the SDSS.

To utilize spatial and nonspatial data, the expert system shell has an interface with external database management system (DBMS) such as GIS, relational databases, and remotely sensed information systems. The communication between the expert system shell and the DBMS can be carried out by intelligent expert systemdatabase communication methods.

To facilitate the utilization of externally stored procedural knowledge such as algorithms, statistical procedures, and mathematical models, an interface with a model base management system (MBMS) should be incorporated into the expert system shell. Parallel to DBMS, MBMS organizes procedural models into an easy to use structure. Calls to MBMS can be invoked by meta knowledge in the expert system shell.

In addition to linkages to DBMS and MBMS, friendly user interface and knowledge acquisition modules are essential parts of the expert system shell for human–machine interaction. Communication between DBMS and MBMS, users, and experts should also be considered. Based on the general architecture, an SDSS shell has been developed, implemented, and applied to construct SDSS to assist spatial decision making tasks. The FLESS is a general SDSS development environment constructed for the purpose of building SDSS to solve specific spatial problems in an effective and efficient manner. To facilitate learning by examples, automatic knowledge acquisition functionalities materialized as the neural network and genetic algorithm modules have been built into the system. Whenever necessary, neural networks and/or genetic algorithms may be utilized by FLESS to extract spatial features or rules from data which can subsequently be stored as a knowledge base for some domain specific problems. Of course, neural networks and genetic algorithms can also be utilized directly to calibrate models or to process data in GIS.

Over the years, a number of SDSSs have been developed for solving practical problems. Typical examples are decision support system for the management of water pollution in tidal river networks, classification systems for remotely sensed images based on fuzzy logic, neural networks, genetic algorithms, and hybrids, as well as system for flood simulation and damage assessment. All of these systems are built fully or partly by the generic SDSS shell discussed above, and they all integrate spatial and nonspatial data, as well as structured and/or unstructured geographical knowledge. Though the applications are problem and site specific, the knowledge and technology are transportable and can be made useful to other places of the world.

Similar to the above approach, artificial intelligence has been employed to construct expert systems for landfill design and management, environmental management, landslide hazard monitoring, geographical database management, and spatial analysis.


In the future, geographical analysis will increasingly rely on information and the associated technology. We will have to find spatial structures and processes from very large databases which are multisourced, multiscaled, heterogeneous, imperfect, and dynamic. Creating spatial data mining and knowledge discovery capability is thus a necessity in the development of intelligent SDSS. Through the electronic super highway, information and knowledge will take on a highly distributive form. The globalization of information and decision making is rapidly taking place. How to discover useful and nontrivial knowledge in such environment is thus crucial in this knowledge age. Effective utilization and integration of information from various sources with geographical knowledge acquired through our in depth analysis of the human and physical processes in space and time will greatly enhance our understanding of the ever changing environment and chart out courses of action for the sustainable development of human race. Artificial intelligence, in cooperation with geographical knowledge, will continue to play a major role in the development of intelligent systems to support our spatial decision making process.