Machine Learning: The Automation of Knowledge Acquition Using Kohonen Self-Organising Map Neural Network

Abstract

In machine learning, a key aspect is the acquisition of knowledge. As problems become more complex, and experts become scarce, the manual extraction of knowledge becomes very difficult. Hence, it is important that the task of knowledge acquisition be automated. This paper proposes a novel method that integrates neural network and expert system paradigms to produce an automated knowledge acquisition system. A rule-generation algorithm is proposed, whereby symbolic rules are generated from a neural network that has been trained by an unsupervised Kohonen self-organising map (KSOM) learning algorithm. The generated rules are evaluated and verified using an expert system inference engine. To demonstrate the applicability of the proposed method to real-world problems, a case study in medical diagnosis is presented.