An Efficient Rules Induction Algorithm for Rough Set Classification

Abstract

The theory of rough set provides a formal tool for knowledge discovery from imprecise and incomplete data. Inducing rules from datasets is one of the main tasks in rough set based data mining. According to Occam Principle, the most ideal decision rules should be the simplest ones. Unfortunately, induction of minimal decision rules turns out to be a NP-hard problem. In this paper, we propose an heuristic minimal decision rules induction algorithm Rulelndu whose time complexity is O(|A|*|U|2) and space requirement is O(|A|*|U|). In order to investigate the efficiency of proposed algorithm, we provide the comparison between our algorithm Rulelndu and some other rules induction algorithms on some problems from UCI repository. In most cases, our algorithm Rulelndu outmatches some other rules induction algorithms not only in classification time but also in classification accuracy. © Springer-Verlag 2004.