The main porpouse of learning is to use the learned rules for classification of new instances. YAILS has several mechanisms to enhance its performance on such tasks. The validity of these mechanisms was tested on several real world domains as well as artificial ones [Torgo,1992]. Nevertheless it's the author's believe that further improvements should be possible namely on the issue of conflict resolution.
The classification module included in YAILS has great flexibility in the sense that it is possible to use different strategies, each one with several parameters. This means that a period of tuning could be needed in face of a new problem in order to obtain the best results.