Systems for inducing classification rules from databases are valuable tools for assisting in the task of knowledge acquisition for expert systems. In this paper, we introduce an approach for extracting knowledge from databases in the form of inductive rules. We develop an information theoretic measure which is used as a criteria for selecting the rules generated from databases. To reduce the complexity of rule generation, the boundary of the information measure is analyzed and used to prune the search space of hypothesis. The system is implemented and tested on some well known machine learning databases.