A flexible knowledge discovery system using genetic programming and logic grammars

As the computing world moves from the information age into the knowledge-base age, it is beneficial to induce knowledge from the information superhighway formed from the Internet and intranet. Knowledge discovery in databases is defined as the non-trivial process of identifying valid, novel, potentially useful, and ultimately understandable patterns in data stored in databases. The knowledge acquired can be expressed in different knowledge representations such as computer programs, first-order logical relations, or Fuzzy Petri Nets (FPNs). In this paper, we present a flexible knowledge discovery system called LOGENPRO (The LOgic grammar based GENetic PROgramming system) that applies genetic programming and logic grammars to learn knowledge in various knowledge representation formalisms. The system is also powerful enough to represent context-sensitive information and domain-dependent knowledge. An experiment is performed to demonstrate that LOGENPRO can discover knowledge represented in FPNs that support fuzzy and approximate reasoning. To evaluate the performance of LOGENPRO in producing good FPNs, the classification accuracy of the fuzzy Petri net induced by LOGENPRO and that of the decision tree generated by C4.5 are compared. Moreover, the performance of LOGENPRO in inducing logic programs from noisy examples is evaluated. A detailed comparison to FOIL, a system that induces logic programs, has been conducted. These experiments demonstrate that LOGENPRO is a promising alternative to other knowledge discovery systems and sometimes is superior for handling noisy and inexact data. Area: Knowledge Discovery in Databases, Genetic Programming, Logic Grammars, Fuzzy Petri Nets