Improved Fuzzy Rough Set model based on Automatic Noisy Sample Detection
Abdelkhalek Hadrani, Karim Guennoun, Rachid Saadane, Mohammed Wahbi · 2019
Rough Set and Fuzzy Set theories have been combined to give birth to a third hybrid approach, named Fuzzy Rough Set (FRS) theory, which is useful in analyzing and processing vague, incomplete and uncertain information. However, several works proved that its classical knowledge representation model has a major weakness related to its sensitivity to data noise which reduces both its efficiency and application scope. This fact motivated us to develop a new FRS model based on the Automatic Noisy Sample Detection (ANSD-FRS) so as to eliminate noise influence in classification tasks. In this paper, we present this novel paradigm, study its main properties and redefine some useful FRS concepts relied on its operators. We also give and discuss experimental results showing the efficiency and robustness of the ANSDS-FRS compared to the most widely used and cited models that are known to be resilient against noise.