A hybrid approach to rule discovery in databases

This paper introduces a hybrid approach for rule discovery in databases in an environment with uncertainty and incompleteness. We first create an appr opriate relationship between deductive reasoning and stochastic process, an d extend the relationship for including abduction. Then, we define a Genera lization Distribution Table (GDT), which is a variant of transition matrix in stochastic process, as a hypothesis search space for generalization, and describe that the GDT can be represented by knowledge-oriented networks. F urthermore, we describe a discovery process based on the network representa tion. Finally, we introduce some extension for making our approach more use ful, and discuss some problems for real applications. We discuss inductive methods from the viewpoint of the value of information, and describe that t he main features of our approach are: (1) the uncertainty of a rule, includ ing its ability to predict possible instances, can be explicitly represente d in the strength of the rule, (2) noisy data and data change can be handle d effectively, (3) biases can be flexibly selected and background knowledge can be used in the discovery process for constraint and search control, an d (4) if-then rules can be discovered in an evolutionary, parallel-distribu ted cooperative mode. (C) 2000 Elsevier Science Inc. All rights reserved.