Transcript Ferda - Knowledge Engineering Group

Ferda
New visual data mining
environment
Martin Ralbovský
LISp-Miner





Academic software project for research
and teaching of knowledge discovery in
databases at VŠE
Uses association rule data mining, GUHA
method
Research and development since 1996,
creation and implementation of new
procedure
9 data mining procedures up-to-date
Complex, but user-challenging tool
Ferda
Software project at MFF UK
 New user interface for LISp-Miner
procedures
 Development since 2004


Created by: Michal Kováč, Tomáš Kuchař,
Alexander Kuzmin, Martin Ralbovský
Main goals of Ferda




To simplify and clarify the process of data
mining with the aid of visual programming
Modularity, extensibility, distributibility
Creation of a strong tool, features to take
the LISp-Miner “one step further”
Environment not only for association rules
Working environment – box





Box is an element of visual programming
in Ferda
Box contains sockets and properties
In comparison to other DM systems, box
is an abstraction of a function rather than
a visual part of DM process
User creates the task by connecting the
boxes and setting their properties
More perspectives on box – programmer
sees programming language, logician sees
formulas
Parts of the environment

Desktops
More desktops for one project

Property grid
Standard component for setting properties

Archive
Storeroom for all the boxes in the project

Docking
Working with boxes
Setting properties with property grid
(possibly other modules)
 Modules for interaction
 Actions
 Boxes asking for creation
 Packing and unpacking of boxes

Archive
Contains all the boxes in a project
 Horizontal browsing along and
against the connections
 Vertical browsing according to box
categories
 Actions supporting desktop and
archive synchronization
Boxes expanding the system
functionality

Basic data types boxes
each property is a hidden socket

Group box
joining everything to groups

Dynamic and not-dynamic attributes
when input data changes with time

Equivalence class
also over attribute
Solving a practical task
…
Ferda – future
Master theses of Ferda creators should
add new functionality
 Generalization of task definitions
with the aid of logic (EverMiner)
 Support for ontology
 Broadening of current procedures to
relational equivalents