keyword

AQ algorithm

The AQ algorithm is a supervised machine learning method designed for inductive concept learning that generates logical classification rules from labeled training examples. Developed as part of a family of rule induction techniques based on a covering strategy, the algorithm produces decision rules in a symbolic format, such as disjunctive normal form. It operates iteratively by selecting an uncovered positive training example as a seed and generating a star, which is a set of maximally general condition descriptions that cover the seed while excluding negative examples. From this star, the algorithm selects the optimal rule based on predefined preference criteria, adds it to the learned rule set, removes all positive instances satisfied by that rule, and repeats the procedure until all positive examples are covered. This approach produces structured, human-interpretable production rules that are widely utilized in expert systems, pattern recognition, and automated knowledge acquisition.

1 item