Soft Techniques for Rule Discovery in Data

概要

論文の詳細を見る
This paper introduces a new approach to rule discovery in data with uncertainty and incompleteness, using soft techniques. This approach is based on a Generalization Distribution Table(GDT)and its network representation. Inductive learning methods are discussed from the viewpoint of the value of information. The key features of the approach are as follows:(1)both the formal value and the semantic value of information are considered, (2)the uncertainty of a rule, including its ability to predict possible instances, can be explicitly represented in the strength of the rule, and(3)biases can be flexibly selected and background knowledge can be used in the discovery process for constraint and search control.
1998-09-15