Přehled o publikaci

Credit risk assessment, credit scoring or loan applications approval are one of the typical classification tasks, in which the final decision can be either a crisp yes/no decision about the loan or a numeric score expressing the financial standing of the applicant. A corresponding classifier can be created from data about past decisions. Beside a logistic regression, that constitutes a de-facto banking industry standard and a benchmark algorithm, a number of data mining and machine learning algorithms can be used as well. In the paper we focus on tree building algorithms. Due to their understandability, the trees can be used not only for classification but also for concept description. Another advantage is that trees can be created also from data with missing values. We present the basic concept of learning trees from data, describe our method for creating exploration trees and discuss its difference with algorithms for creating decision trees. We also compare our method with a standard tree learning algorithms C4.5 and CART on some data from the loan application domain.