Rough problem settings for ILP dealing with imperfect data

This paper applies rough set theory to Inductive Logic Programming (I LP, a relatively new method in machine learning) to deal with imperfect data occ urring in large real-world applications. We investigate various kinds of im perfect data in ILP and propose rough problem settings to deal with incompl ete background knowledge (where essential predicates/clauses are missing), indiscernible data (where some examples belong to both sets of positive and negative training examples), missing classification (where some examples a rc unclassified) and too strong declarative bias (hence the failure in sear ching for solutions). The rough problem settings relax the strict requireme nts in the standard normal problem setting for ILP, so that rough but usefu l hypotheses can be induced from imperfect data. We give simple measures of learning quality for the rough problem settings. For other kinds of imperf ect data (noise data, too sparse data, missing values, real-valued data, et c.), while referring to their traditional handling techniques, we also poin t out the possibility of new methods based on rough set theory.