Seminář: User oriented language for powerful data mining with the Ferda Data Miner, Implementation of GUHA decision trees – initial remarks

The Ferda Data Miner has been developed as user oriented data mining tool. User experience is strong part of the Ferda system. The data mining tasks so far consisted only of individual runs of GUHA procedures implemented in the LISp-Miner system and other procedures implemented in the system. These tasks were not comparable with data mining tasks in tools like MATLAB in terms of task variability. Main power in tools like MATLAB is it’s programming language. New programming language is in developement for the Ferda. With aid of this language, one can create user defined data mining tasks. The language is in many ways different from a standard programming language. The code is not written as a text but as a connection of boxes. It is functional language, but the main strength of it is in special version of lambda pattern, which is not used in other functional languages. The new functional language for the Ferda Data Miner will be shown with examples for data mining.

The GUHA method provides a general mainframe for retrieving interesting information from data and has been used mainly for enhanced association mining. Yet the mainframe can be used for other forms of data mining. Petr Berka suggested a GUHA – like decision tree mining procedure named ETree. The method has recently been implemented in the Ferda system. Theoretical basics of the procedure will be presented as well as implementation details.