Background knowledge consists both of domain-specific knowledge as well as general knowledge about the behavior of the world. Use of background knowledge in the process of identifying general patterns within a database leads to patterns that are more useful and significant. The purpose of this project is to facilitate extensive usage of background knowledge about the domain during the process of extracting rule-based descriptions from large databases. The project will include both a comparison of existing tools for the incorporation of background knowledge such as FOIL, C4.5, and Golem, as well as the design and Prolog language implementation of a new system.
LEARNING FIRST ORDER RULES FROM EXAMPLES
- Faculty: Sergio A. Alvarez, Carolina Ruiz.
- Students: Matt Berube.
Project DescriptionWe have developed a learning prototype called TinFOIL, based on Quinlan's FOIL algorithm. The TinFOIL system distills first-order logic rules from examples and can be used to mine data from large databases. We have experimented with applying TinFOIL to several data sets, including the Income Census Dataset from the University of California, Irvine machine learning repository.