Philippe Levecq

A comparison of rough sets and recursive partitioning induction approaches : an application to commercial loans

Credit scoring is the term used to describe methods utilised for classifying applicants for credit into classes of risk. This paper evaluates two induction approaches, rough sets and decision trees, as techniques for classifying credit (business) applicants. Inductive learning methods, like rough sets and decision trees, have better knowledge representational structure than neural networks or statistical procedures because they can be used to derive production rules. If decision tree have already been used for credit granting, the rough sets approach is rarely utilised in this domain. In this paper, we use production rules obtained on a sample of 1102 business loans in order to compare the classification abilities of the two techniques. It results that decision tree obtains better results with 87.5% of good classifications with a pruned tree against 76.7% for rough sets. However, decision tree make more type II errors than rough sets but less type I errors.

Multivariate Analysis and Rough Sets: Two Approaches for Software‐quality Analysis

Computer programs are becoming more and more complex and voluminous in most firms. This raises the cost of maintenance. To minimize this cost, qualities must be given to computer programs that assure them good evolution, i.e., reliability, efficiency, flexibility, conformity, etc. This cost minimization must be preceded by a diagnosis that recommend a set of renovation actions to be undertaken in order to improve the software quality. The objective of this study is to associate a set of renovation action to each profile’s program expressed in terms of metrics. To do this, two approaches are taken. First, a multivariate analysis is used to allocate each program to a predefined category reflecting the diagnosis of the program’s health. Second, we try to work out a diagnosis of the health of these programs by developing a quality model composed of a set of decision rules supporting software quality improvement decisions. We have finally implemented this set of rules in the form of an expert system tool to allow the classification of new programs by the decision-maker.

How Can Help an Inductive Approach in the Resolution of a Problematic \alpha Problem?

Some organizations are commissioned to boost every kind of innovation in favoring collaboration between small businesses and scientific circles. Our research aims at developing a decision-aid tool to help intermediate organizations in their search for innovative enterprises and at determining if these enterprises are receptive to collaboration with a university. This helping tool consists of a set of decision rules thanks to which the enterprises are selected. This set of rules was established with the rough set method. The problem we have to face comes under the problematic P.a and the originality of the paper is that we will show the impact of the choice of decision rules on the type I error and thus on the percentage of objects incorrectly classified.