What criteria can be used to decide whether to develop a decision tree or one or more rules using oneR or jRip(contained in RWeka). The similarities between these two algorithms (decision tree and rule) is very high making it a little confusing.
$\begingroup$
$\endgroup$
4
-
$\begingroup$ What are oneR and jRip? Haven't heard them before. Can you explain a bit and add some details to your question? $\endgroup$– Dawny33Apr 26, 2016 at 11:42
-
$\begingroup$ These are functions in R. OneR makes a single classification rule. While jRip will make as many rules as necessary $\endgroup$– Darrin ThomasApr 26, 2016 at 11:47
-
$\begingroup$ Let me know if my edit helps? $\endgroup$– Dawny33Apr 26, 2016 at 11:51
-
$\begingroup$ Cool. No problem $\endgroup$– Darrin ThomasApr 26, 2016 at 12:01
Add a comment
|
1 Answer
$\begingroup$
$\endgroup$
JRip implements a propositional rule learner, “Repeated Incremental Pruning to Produce Error Reduction” (RIPPER), as proposed by Cohen (1995) and OneR builds a simple 1-R classifier, proposed by Holte (1993).
Its hard to say which algorithm works better. Best approach is to compare different classification algorithms performance in terms of precision, recall, accuracy, f1 score, AUC, specificity and sensitivity on your train/test data set and pick the one that gives best results or use ensemble of top performing algorithms to build your final model.
There is a good white paper doing similar exercise of comparing different classification algorithms including OneR and JRIP here.
Hope this helps.
answered Apr 26, 2016 at 13:18