Knowledge Acquisition for Spatial Inference Using Genetic Algorithms

Principal Investigator:
Leung, Yee

Co-investigator(s):
Leung, Kwong Sak

Summary:
Intelligent spatial decision support systems integrating human expertise and very large spatial databases have become a necessity for solving our highly complex spatial problems such as resource exploration, land-type classification, transportation planning, and environmental management. In addition to a sound theoretical foundation, a good system design, and a powerful and user-friendly software environment, the success of intelligent decision support systems lies heavily on the domain specific knowledge they acquired. Knowledge acquisition thus plays a very important role in machine intelligence.

The purpose of this project is to develop an intelligent self-learning system in spatial decision support systems using genetic algorithms (GAs) so that knowledge can be automatically acquired to solve complicated problems involving voluminous geographic and remotely sensed information. Differing from conventional GAs, the proposed algorithm will employ a highly compressed representation method to structure fuzzy and non-fuzzy rules for efficient self-learning by examples. Two novel features, namely token competition and rule migration, will be incorporated in the GA to accelerate learning through maximum diversity.

The project will advance a research frontier in machine learning in spatial decision-making problems using GAs. It will pave the road for solving the bottleneck of knowledge acquisition in large-scale spatial inference involving geographic information and remote sensing systems. The powerful and user-friendly software environment will be an effective decision support tool for researchers and policy makers to solve local and regional spatial problems.