Towards a Logic Query Language for Data Mining

We present a logic database language with elementary data mining mechanisms to model the relevant aspects of knowledge discovery, and to provide a support for both the iterative and interactive features of the knowledge discovery process. We adopt the notion of user-defined aggregate to model typical data mining tasks as operations unveiling unseen knowledge. We illustrate the use of aggregates to model specific data mining tasks, such as frequent pattern discovery, classification, data discretization and clustering, and show how the resulting data mining query language allows the modeling of typical steps of the knowledge discovery process, that range from data preparation to knowledge extraction and evaluation.