An Empirical Study of Using Rule Induction and Rough Sets to Software Cost Estimation

Issue title: Theory and Applications of Soft Computing (TASC04)

Article type: Research Article

Authors: Stefanowski, Jerzy

Affiliations: Institute of Computing Sciences, Poznań University of Technology, ul. Piotrowo 3a, 60-965 Poznań Poland. E-mail: Jerzy.Stefanowski@cs.put.poznan.pl

Abstract: This paper concerns problems of applying the approach based on rough sets and rule induction to a software engineering data analysis. More precisely, we focus our interest on a software cost estimation problem, which includes predicting the effort required to develop a software system basing on values of cost factors. The case study of analysing the COCOMO data set, containing descriptions of representative historical projects, allows us to discuss how this approach could be used to: identify the most discriminatory cost factors, extract meaningful rule representation of classification knowledge from data, construct accurate rule based classifiers.

Keywords: Software Engineering, Rough Sets, Machine Learning, Preference Modeling, Software Cost Prediction, COCOMO model

Journal: Fundamenta Informaticae, vol. 71, no. 1, pp. 63-82, 2006