Induction and ID/3: more powerful than we think
Daniel Groß · Expert Systems · 1988
Abstract: Most existing expert systems store rules which guide programme behaviour. Inductive expert systems generate rules from properly formatted descriptions of historical data. The ID/3 algorithm, based on Earl Hunt's Concept Learning System, is among the more popular induction algorithms for this application. A general description of ID/3 is given along with suggestions for further possible refinements of the algorithm. Because ID/3 uses decision trees to store knowledge and does not inherently exclude the application of weights to paths through the tree, the algorithm bears a remarkable conceptual resemblance to neural networks. From this analogy some new applications for the algorithm are derived.