Knowledge Discovery in Architectural Cad Data

Knowledge discovery and data mining techniques have the potential to find patterns in very large sets of data. Applying these techniques to design data has two problems: the volume and complexity of the source data, and the discovered knowledge itself. A result is that existing algorithms may not be computationally feasible and in any case may not produce, of themselves, results that are of interest to a designer. In this paper we first demonstrate the two problems above by applying data mining techniques to architectural CAD data. We then reconsider the role of data mining for knowledge discovery as facilitating the re-presentation of source design data in other forms such that relationships are revealed to the designer that would not otherwise be visible. To investigate this we adopted a framework for knowledge discovery combining consecutive complementary strategies, and applied it to the same architectural CAD data.