Data Mining Using Multi-Valued Logic Minimization

In a partially defined classification function, each input combination represents features of an example, while the output represents the class of the example. Each variable may have different radix. In this paper, we show a method to minimize the number of variables. Combined with a multiplevalued logic minimizer, data sets of examples are represented by a compact set of rules. Experimental results using University of California Irvine (UCI) benchmark functions show the effectiveness of the approach, especially for imbalanced data sets. The results are compared with J48 and JRIP. This approach produces explainable 100% correct rules, which are promising for bio-medical applications.