A New Hybrid Architecture for the Discovery and Compaction of Knowledge from Breast Cancer Datasets

This paper reports on a two-fold contribution; first, the introduction of a new compaction algorithm for the rules generated by learning classifier systems that overcomes the disadvantages of previous algorithms in complexity, compacted solution size, accuracy and usability. The second is the new hybrid architecture that integrates learning classifier systems with Rete-based Inference Engines to improve the performance of extracting a minimal and representative ruleset from the original learning classifier systems generated ruleset, when applied to a breast cancer pathological dataset. In addition to demonstrating significant savings in computing the match phase, this has resulted in enhancing the readability of the generated rules for subsequent validation of the generated knowledge by domain experts. Finally, this hybrid architecture that is component-based, and extensible, establishes a new platform for research on the efficiency and rules compaction of learning classifier systems using Rete-based inference engines.