This thesis addresses the problem of incorporating uncertainty in
attribute values when building classification models in the context of machine
learning. Traditional algorithms, such as decision trees and random forests,
typically ignore uncertainty in attribute values, which can negatively impact
the reliability of results. As part of this work, I implemented custom widgets
for the Orange data mining software, enabling the use of extended versions of
these algorithms that account for uncertainty. This allows for visual analysis
and comparison of results in a user-friendly environment, accessible even to
less experienced users. The main contribution of the thesis is the extension
of Orange’s functionality and a foundation for further research in this area.
|
