NONDETERMINISTIC DECISION RULES

Abstrakt

In the paper the authors discuss classifiers based on deterministic decision rules and non-deterministic decision rules and prove that non-deterministic decision rules can be used for improving the quality of classification. The authors propose classifications algorithms based on non-deterministic rules and minimal rules in the sense of rough sets. The classifiers in question are tested on the group of decision tables from the UCI Machine Learning Repository and the results are compared. The reported results of experiments show that proposed classifiers based on non-deterministic rules give the possibility to improve the classification quality but with the requirement of tuning to the data.