Decision Tree Classifiers Based on Granular Computing

Bottle-up and top-down are two main computing models in granular computing (GrC). The bottle-up granular computing is used to form decision tree classifiers, or DTCGrC for short. Algorithm DTCGrC constructs a framework of granular computing by the bottle-up join operation which maps all the training data into the granule set, and the achieved granule set is used to form the decision tree classifiers. We compare the performance of DTCGrC with decision tree classifiers, for a number of two-class problems and multiclass problems. Our computational experiments showed that DTCGrC improves the generalization abilities.