| Classication Rule Induction from Databases |
|
The primary objective of this research is to develop a learning algorithm for rule set induction
using second-order relations as a new representational model. Second-order relations are database
relations in which tuples, which represent rules, have sets of atomic values as components. Using
sets of values, which are interpreted as disjunctions, provides a compact representation that both
facilitates efficient data management and enhances comprehensibility. This representation is
intuitively appealing for dealing with large databases, especially those of highly structured knowledge.
By enhancing the scalability of the learning approach with respect to both complexity of data (as measured
by the number of attributes and attribute values involved) and the size of the training set, this research
will extend the learning algorithm to a data mining algorithm for classification that scales to large databases.
|
Secondary objectives include investigations of how well our relational framework and learning systems
- Facilitate adequate and efficient information extraction
- Facilitate integration with relational database systems
- Allow incorporation of prior knowledge
- Support interactive and incremental learning
|
| Related keywords: Data Mining, Machine Learning, Classification,
Incremental Learning, Decision Support Systems, Relational Database Theories,
Boolean Minimization
|
| Collaborators: John Leuchner, David Collodi
|
|
Mailing address: Center for Excellence in Engineering Graduate Studies and Research, 302 Pine St. Abilene, TX 79601. Tel: (325) 677-1112 or
Department of Computer Science, P.O. Box 43104, Lubbock, TX 79409-3104. Tel: (806)742-3527. |
|
