A three-phase knowledge extraction methodology using learning classifier system

Machine learning methods such as fuzzy logic, neural networks and decision tree induction have been applied to learn rules but they may be trapped into local optimal. Based on the principle of natural evolution and global searching, a genetic algorithm is promising in obtaining better results. This article adopts learning classifier systems (LCS) technique to provide a three-phase knowledge extraction methodology, which makes continues and instant learning while integrates multiple rule sets into a centralized knowledge base. This paper makes three important contributions: (1) it represents various rule sets that are derived from different sources and encoded as a fixed-length bit string in the knowledge encoding phase; (2) it uses three criteria (accuracy, coverage, and fitness) to select an optimal set of rules from a large population in the knowledge extraction phase; (3) it applies genetic operations to generate optimal rule sets in the knowledge integration phase. The experiments prove the rule sets derived by the proposed approach is more accurate than other machine learning algorithm.