A hill-climbing algorithm combined with fuzzy set concepts toward the development of a compact associative classifier

Classification being one of the most frequently-used data-mining techniques has widespread applications including credit approval, medical diagnosis and targeted marketing. Knowledge discovery from databases in the form of association rules is an important data-mining task. Several algorithms and methods have been devised over the past decade toward building associative classifiers that could achieve high classification accuracies. While attention has been mainly focused on increasing classifier accuracies, not much effort has been devoted toward building interpretable models that have small number of classification rules. Classification accuracy and model interpretability tend to act against one another. That is, a larger rule-base generally provides higher classification accuracy and a smaller rule-base tends to yield lower classification accuracy.;This research develops an associative classification model using the hill-climbing approach and the concept of fuzzy sets. The proposed methodology builds the rule-base by selecting rules which contribute toward increasing training accuracy, thus balancing classification accuracy with the number of classification association rules. The results indicated that the proposed associative classification model can achieve competitive accuracies on benchmark data sets with continuous attributes when compared to other rule-based systems. The compact rule-bases lend better interpretability to the proposed classification models.