Abstract
It is important to discover hierarchical decision rules from databases because much of the world’s knowledge is best expressed in the form of hierarchies. Mining of decision rules at multiple concept levels leads to discovery of more informative and comprehensible knowledge. This paper proposes automated discovery of Hierarchical Production Rules (HPR) using a parallel genetic algorithm approach. A combination of degree of subsumption and coefficient of similarity has been used as a quantitative measure of hierarchical relationship among the classes. An island/deme GA is designed to evolve HPRs for the classes of the dataset being mined. The island model exploits control as well as data parallelism. The model is applied to a synthetic dataset on means of transport and results are presented.
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call