keyword

CN2 induction algorithm

The CN2 induction algorithm is a machine learning technique designed to automatically generate simple and comprehensible classification rules from labeled training data. Developed to perform robustly in the presence of noise and imperfect descriptions, it combines the beam search and sequential covering strategies of the AQ family of algorithms with the statistical significance testing and information-theoretic evaluation criteria found in decision tree algorithms like ID3. Instead of constructing a full decision tree, CN2 iteratively searches for optimal if-then rules that cover subsets of training examples, removes the covered instances from the dataset, and repeats this process until a complete rule set or decision list is produced for classifying new, unseen examples.

1 item