Mass assignment-based induction of decision trees on words
JF Baldwin, Jonathan Lawry, T. Patrick Martin · 1998
A mass assignment based ID3 algorithm for the induction of decision trees on words is described. Such decision trees encode sets of qualified conditional rules on linguistic variables. The potential of this algorithm is illustrated by means of several examples relating to both real world and model classification and prediction problems. Keywords:: mass assignment, decision tree, linguistic variable, induction 1 Introduction The ID3 and C4.5 algorithms (see [9] and [11]) have been successfully applied to a wide variety of machine learning problems. It is well known, however, that such approaches have some limitations. For instance, ID3 is inappropriate for databases containing significant noise since the generated rules will fit the noise which may lead to a high error rate when classifying unseen cases. Furthermore, often in practice classification problems have continuous attribute values associated with them necessitating the partitioning of relevant universes if ID3 type algorithm...