Recent Advances in Intelligent Information Systems ISBN 978-83-60434-59-8, pages 567–578 Inference Algorithms for Hierarchical Knowledge Bases Agnieszka Nowak, Roman Simiński, and Alicja Wakulicz-Deja Institute of Computer Science, University of Silesia 39, Będzińska St., 41-200 Sosnowiec, Poland, http://zsi.tech.us.edu.pl agnieszka.nowak,roman.siminski,wakulicz@us.edu.pl Abstract In this paper, we present new approach of inference processes for large, complex knowledge bases. Nowadays, knowledge bases characterized by complex inner rules structure and by huge number of rules. Such bases can count up to hundreds or thousands of rules, what causes problems not only with inference efficiency but also with interpretation of inference results. We describe both methods of creating hierarchically organized knowledge bases: cluster analysis method — for forward inference and decision units method — for backward inference. The experimental results done on a real-world rule bases show that proposed inference algorithms improving the efficiency of inference especially for large rule bases. We also present conclusions and future works goals. Keywords: rule knowledge bases, inference algorithms, clustering, decision units 1 Introduction The very well known fact about knowledge based systems technology in recent years is that it is a valuable tool for solving problems in many domains of compe- tence, in which effective problem solving normally requires human expertise. The reasons of building such systems are following: they are often more ambitious than of conventional programs, they frequently perform not only as problem solvers but also as intelligent assistant and training aids and they help their creators and users to understand better own knowledge(Luger , 2002). Knowledge based systems are usually implemented using knowledge acquired from human experts or discovered in data bases (for example using rough set theory (Skowron et al., 2007) and methods of construction partial decision rules (Moshkov et al., 2008b)). The most popular method of knowledge representation is still the rule based method. Unfortunately if we use — possibly different — tools for automatic rules acquisition and/or extraction, the number of rule can rapidly grow. For modern problems knowledge bases can count up to hundreds or thousands of rules, what causes problems not only with inference efficiency but also with interpretation of inference results. For example, if we consider forward reasoning, a lot of fired rules forming a lot of new facts that are sometimes difficult