| Abstract | This paper discusses the advantages of knowledge-driven constructive induction (KDCI). Development, testing, and evaluation of a KDCI (or in short CI for constructive induction) system are explained in detail. The objectives of developing this system were to demonstrate theusefulness of the approach and to provide knowledge-driven constructiveinduction support in our data analysis research. Technical details, particularly the process of building new attributes and changing the representation space, are discussed. Other issues concerning the design and implementation of the CI mechanism are presented. The evaluation process and comparison measures used for evaluation of the system are briefly explained. Experimental results, using 4 data sets from a real-world application, are given. |
|---|
