A Probabilistic Approach to Classify Incomplete Objects Using Decision Trees

Speaker: TSANG Pui Kwan

Abstract

We describe an approach to fill missing values in decision trees during classification. This approach is derived from the ordered attribute trees method, proposed by Lobo and Numao in 2000, which builds a decision tree for each attribute and uses these trees to fill the missing attribute values. It is based on the Mutual Information between the attributes and the class. Our approach primarily extends this method on three points: 1) it does not impose an order of construction; 2) a probability distribution is used for each missing attribute instead of the most probable value; 3) the result of the classification process is a probability distribution instead of a single class. Moreover, our method takes the dependence between attributes into account. We present Lobo's approach and our extensions, we compare them, and we discuss some perspectives.