Abstract
Conventional Fuzzy Classification Rules (FCRs) are most often discovered in the form of flat ‘If-Then’ rules. These flat rules increase the size of Fuzzy Rule Based Systems (FRBSs). Moreover, a large number of rules are not considered comprehensible. Most of the real world knowledge can be organized into hierarchical fashion for deciphering knowledge at multiple levels of details. However, the flat representation of rules ignores hierarchical relationships that may exist among the classes of a dataset. This paper proposes a Genetic Algorithm approach to discover Fuzzy Hierarchical Classification Rules (FHCRs) . The rules discovered in this form can predict the knowledge at various level of abstraction. As a part of the GA design, a suitable encoding scheme to capture the hierarchical structure of knowledge being mined, a fitness function to measure goodness of the hierarchies and genetic operators to evolve the solutions have been suggested. The proposed approach is illustrated on a ‘Land transport’ dataset specifically designed for the purpose.
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