A Multi-Relational Rule Discovery System.pdf

A Multi-Relational Rule Discovery System Mahmut Uluda?1, Mehmet R. Tolun2, Thure Etzold1 1 LION Bioscience Ltd., Compass House, 80-82, Newmarket Road, Cambridge, CB5 8DZ, United Kingdom {mahmut.uludag, thure.etzold}@ 2 Atilim University, Dept. of Computer Engineering, 06836 Incek, Ankara, Turkey tolun@.tr .tr/rila/ Abstract. This paper describes a rule discovery system that has been developed as part of an ongoing research project. The system allows discovery of multi- relational rules using data from relational databases. The basic assumption of the system is that objects to be analyzed are stored in a set of tables. Multi- relational rules discovered would either be used in predicting an unknown ob- ject attribute value, or they can be used to see the hidden relationship between the objects’ attribute values. The rule discovery system, developed, was de- signed to use data available from any possible ‘connected’ schema where tables concerned are connected by foreign keys. In order to have a reasonable per- formance, the ‘hypotheses search’ algorithm was implemented to allow con- struction of new hypotheses by refining previously constructed hypotheses, thereby avoiding the work of re-computing. 1 Introduction Most of the current data mining algorithms are designed to use data from a single table. They require each object to be described by a fixed set of attributes. Compared to a single table of data, a relational database containing multiple tables makes it pos- sible to represent more complex and structured data. In addition, today, a significant amount of scientific data is stored in relational databases. For these reasons, it is im- portant to have discovery algorithms running for relational data in its natural form without requiring the data to be viewed in a single table. A relational data model consisting of multiple tables may represent several object classes, i.e. within a schema while one set of tables represents a class of objec