Cocomo

JOURNAL OF SOFTWARE, VOL. 3, NO. 5, MAY 2008

65

Empirical validation of Software development effort multipliers of Intermediate COCOMO Model
Basavaraj M.J
Perot Systems, EPIP Phase II, Whitefield Industrial Area, Bangalore-560 066, India Email : basavaraj.m@ps.net; basavarajmj@hotmail.com

Dr. K.C Shet
Professor, Computer Department, National Institute of Technology Karnataka,Surathkal, India Email : kcshet@nitk.ac.in; kcshet@yahoo.co.uk

Abstract—Intermediate COCOMO Model computes effort as a function of program size and a set of cost drivers. Effort adjustment factor (EAF) is calculated using 15 cost drivers. EAF is an important significant factor in computing software development effort. We have taken one delivered development project of size of 479 function points andplanned for 917 Person days of SEI CMM Level 5 “Excellent” Company as a case study to analyze the EAF. We have empirically validated the cost driver model for Intermediate COCOMMO using this projects data. Validation has been done by using other two development projects data of Excellent Company. From our analysis, we have found that cost drivers defined ratings need to be revisited for the projectsof size less than 10 Person months. We have come out with ratings for some cost drivers where earlier it was not defined. This approach helps the project managers to anticipate and estimate the efforts for development projects preferably less than 10 Person months. We have achieved approximately 30% improvement in effort variance by following this approach. Keywords—Software Estimation,Intermediate COCOMO Model

A. Literature Review We have undertaken literature review to study work done till now by others in this context. Improving software effort estimation does not necessarily require adopting sophisticated formal estimation models or expensive project experience databases[2]. Jorgensen argues that estimation using expert judgements are better than models[2]. Fran Niessink and Hansvan Vliet[3] clams that existence of a consistently applied process is an important and a prerequisite for a successful measurement program in case of different environments. There are number of ways to determine the effort needed in software development projects. In traditional software cost models, costs are derived from effort. Empirical estimation models provide computational formulae forcalculating the effort based on statistical approach by referring the past data of more or less similar projects executed[9][10] COCOMO (Boehm, 1981) is the one of the best of these models. Boehm states that COCOMO’s intermediate model gives estimates which varies from the actual needed effort about 20% in average. COCOMO-II (Boehm et al., 2000) is a new updated version of the model, with a moremodern project database[9][10]. The Intermediate COCOMO model computes effort as a function of program size and a set of cost drivers[4]. Usage of cost driver is significant from the point of view of Project Managers while estimating projects which are of less size in person months. Besides, such studies are sparsely available in literature. B. Scope of this Work This paper explains the Empiricalvalidation for software development effort multipliers of Intermediate COCOMO model and analysis has been done to define the ratings for some cost drivers of EAF. Advantages are listed below.

I. INTRODUCTION In the intention of maintaining security and confidentiality of data, authors are constrained not to disclose the company or client name or project name or exact named data in their research. Inthis context Company name “Excellent”, Project name “A”, Project name :”B”, Project name “C” and client name “Super” refer some dummy names. Authors intended to use the past data of SEI CMM Level5 matured company “Excellent” of Project “A”, Project “B” and Project “C” to empirically validate the Intermediate COCOMO Model by predicting cost drivers.

© 2008 ACADEMY PUBLISHER

66

JOURNAL...