We propose an improved inductive learning method to derive classification rules correctly describing (at least) most of the positive examples and do not correctly describe (at least) most of the negative examples. We start with a pre-analysis of data to assign higher weights to those values of attributes
which occur more often in the positive than in the negative examples. The inductive learning problem is represented as a modification of the set covering
problem which is solved by an integer programming based algorithm using elements of a greedy algorithm or a genetic algorithm, for efficiency. The results are very encouraging and are illustrated on a thyroid cancer data set. |