Empirical evaluation of a fuzzy logic-based software quality prediction model
Sun Sup So ∗ , Sung Deok Cha, Yong Rae Kwon
Department of Electrical Engineering & Computer Science (EECS), Korea Advanced Institute of Science and Technology, 373-1, Kusong-dong, Yusong-gu, Taejon 305-701, South Korea Received 14 October1998; received in revised form 17 February 2001; accepted 11 May 2001
Abstract Software inspection, due to its repeated success on industrial applications, has now become an industry standard practice. Recently, researchers began analyzing inspection data to obtain insights on how software processes can be improved. For example, project managers need to identify potentially error-prone softwarecomponents so that limited project resource may be optimally allocated. This paper proposes an automated and fuzzy logic-based approach to satisfy such a need. Fuzzy logic o ers signiÿcant advantages over other approaches due to its ability to naturally represent qualitative aspect of inspection data and apply exible inference rules. In order to empirically evaluate the e ectiveness of our approach,we have analyzed published inspection data and the ones collected from two separate inspection experiments which we had conducted. 2 analysis is applied to statistically demonstrate validity of the proposed quality prediction model. c 2002 Elsevier Science B.V. All rights reserved. Keywords: Software inspection; Quality prediction; Software metrics; Statistical process control; Fuzzy logic;Inspection metric
1. Introduction Software inspection  is widely considered as an essential practice to develop high-quality software in cost-e ective manner. Much of past studies focused on measuring the e ectiveness of inspections,
S.D. Cha is a liated to the Advanced Information Technology Research Center (AITrc) in KAIST. This work was supported, in part, by the Korea Science and EngineeringFoundation (KOSEF) through the telepresence project at the AITrc. ∗ Corresponding author. Tel.: +82-42-869-5558; fax: +82-42869-3510. E-mail address: email@example.com (S.S. So).
improving inspection processes, and empirically comparing e ectiveness of inspection to other software quality assurance techniques [23,24,27]. Recently, however, researchers began applying statisticalanalysis on inspection data so as to improve overall software development processes and, ultimately, software quality and productivity. Christenson et al.  proposed a set of control charts to estimate the number of defects remaining in the inspected products. Information on the number of residual errors can guide project managers in determining which products need to be inspected again and how thecurrent inspection process might be
0165-0114/02/$ - see front matter c 2002 Elsevier Science B.V. All rights reserved. PII: S 0 1 6 5 - 0 1 1 4 ( 0 1 ) 0 0 1 2 8 - 2
S.S. So et al. / Fuzzy Sets and Systems 127 (2002) 199–208
improved. Similarly, Barnard and Price  have deÿned nine key metrics that project managers can use to plan, monitor, and improve inspection processes.Furthermore, they demonstrated inspection data to be useful in assessing the number of residual faults in the code after inspection. Although statistical methods have been extensively used in the area of reliability and quality, they su er from the unrealistic assumption of error distribution and thus often generate unsatisfactory solutions. Errors tend to be clustered rather than being evenlydistributed [5,17]. Project managers, therefore, are usually more concerned about the degree of error-proneness of the modules rather than the precise estimation on the number of residual errors. If it were possible to identify potentially error-prone modules with relatively high degree of accuracy at little or no extra cost by analyzing existing inspection data, project managers could use such ÿndings...