Work

Attribute-Oriented Induction in Relational Databases.

Year: 1991
Type: article
Abstract: It is beneficial as well as challenging to learn knowledge rules from relational databases because of the vast amount of knowledge implied in databases and the large amount of data stored in databases. In this thesis, we develop an attributeoriented induction method to extract characteristic rules and classification rules from relational databases. The method adopts the artificial intelligence from examples paradigm and applies an attribute-oriented concept tree ascending technique in the learning process which integrates database operations with the learning process and provides a simple, efficient way of learning from large databases. Conjunctive rules as well as restricted forms of disjunctive rules are learned using this method. Moreover, by incorporating statistical techniques, qualitative rules with quantitative information can be learned and noisy data and exceptional cases are elegantly handled. Our analysis of the algorithms indicates that attribute-oriented induction substantially reduces the complexity of database learning processes. A prototype database learning system, DBLEARN, has been designed and implemented; early experiments with the prototype system illustrate the promise of attribute-oriented learning in relational databases.
Author: Yandong Cai
Language: English
Open Access status: green
OpenAlex ID: W1484833499
FWCI: 15.09
Cites: 51
Cited by: 235
Related to: 10