Experimental Results
Experiments with the basic version of CFS described
above have shown that it can be of use to machine
learning algorithms in terms of improving accuracy and
comprehensibility of induced models (Hall and Smith,
1998; Hall 1998). In this paper we look at how CFS
compares with a well known wrapper feature selector.
This section present the results of experiments designed
to compare the performance of common machine learning
algorithms after feature selection by CFS with their
performance after feature selection by the wrapper. In
particular, the accuracy of learners and the size of models
produced after feature selection are compared. In
addition, the execution time of CFS is compared with
the wrapper.