Case studies: Public domain, multiple mining tasks systems: ROSETTA rough sets
Jan Komorowski, Aleksander Øhrn, Andrzej Skowron · Oxford University Press eBooks · 2002
Research in rough sets (Pawlak, 1981, 1982) has resulted in a number of software tools for data mining and knowledge discovery from databases (KDD). Among many of these tools, the ROSETTA system (Ohrn, 1999, Ohrn and Komorowski, 1997; Ohrn et al., 1998) is probably one of the most complete software environments for rough set operations. In ROSETTA, the experimental nature of inducing classifiers from data is explicitly maintained by organizing the workspace in a tree structure that displays how input and output data relate to each other. ROSETTA supports the overall KDD process: from browsing and preprocessing of the data, to reduct computation and rule synthesis, to validation and analysis of the generated rules. Learning may be both supervised (resulting in if-then rules) or unsupervised (resulting in general patterns), and input data may be categorical, numerical, or both. ROSETTA is not tied to any particular application domain, and it has been put to use for a variety of tasks. ROSETTA is a cooperative effort between researchers at NTNU in Norway and Warsaw University in Poland, and is available on the World Wide Web (http://www.idi.ntnu.no/ ~aleks/rosetta/). The system runs under Windows NT/98/95/2000.