The Influence of the Drastic T-norm on Two Fuzzy- Rough Rule Induction Algorithms

Algorithms are widely used for knowledge discovery, and due to the variety of different problem types, there is no single algorithm that solves all problems. Therefore, different types of algorithms have been developed and one of these groups is the fuzzy-rough rule algorithms. In this direction, the paper aims to improve the output performance of the algorithm models by introducing a new norm, in this case the Drastic T-norm. This norm is important because it is part of the various key components of the fuzzy rough algorithms, the lower and upper approximations. To examine the influence of the Drastic T-norm on multiple data sets with different numbers of attributes and instances and from different disciplines, we conducted several experiments in which we used four fuzzy tolerance relationship metrics and a comparison with Einstein T-norms. The results of the experimental evaluation came to the conclusion that, in addition to some improvements that the drastic T-norm provided for some data sets, there were also data sets whose models were inferior compared to the previously used T-norm. Building on this, our future work will focus on examining the influence on other elements of the induction process, such as weak measurements and implicators.