RRULES: An improvement of the RULES rule-based classifier
This is an incremental improvement for users of rule-based classifiers, addressing overfitting and computational efficiency in inductive learning algorithms.
The authors tackled the problem of overfitting and inefficiency in the RULES rule-based classifier by introducing RRULES, which optimizes rule detection and stopping conditions, resulting in a more compact rule set with higher test accuracy, reducing coverage rate by up to a factor of 7 and running 2-3 times faster across datasets.
RRULES is presented as an improvement and optimization over RULES, a simple inductive learning algorithm for extracting IF-THEN rules from a set of training examples. RRULES optimizes the algorithm by implementing a more effective mechanism to detect irrelevant rules, at the same time that checks the stopping conditions more often. This results in a more compact rule set containing more general rules which prevent overfitting the training set and obtain a higher test accuracy. Moreover, the results show that RRULES outperforms the original algorithm by reducing the coverage rate up to a factor of 7 while running twice or three times faster consistently over several datasets.