Abstract:
Knowledge Discovery in Databases (KDD)
processes are complex, highly interactive and iterative. The
similarities between KDD processes and software
development processes suggest that approaches used to
manage the development of software processes are also
applicable and in fact advantageous to KDD processes. In this
paper, we examine the current approaches for supporting
KDD and note to their limitations in providing comprehensive
and effective process support. We propose a language-based
and process-oriented approach for supporting KDD processes
that is based on explicitly representing KDD processes as
process programs that can be analyzed, validated, and
enacted. We illustrate the proposed approach using a novel
process programming language that is designed to describe
general process concepts as well as specific KDD concepts.
Along with the KDD process language, an IDE-style
development environment is proposed to assist in modeling
and enacting KDD processes. The overall approach is
evaluated and illustrated by modeling and enacting a
traditional KDD process.
