by Michael Kläs, Adam Trendowicz, Axel Wickenkamp, Jürgen Münch, Nahomi Kikuchi, Yasushi Ishigai
Abstract:
Cost estimation is a crucial field for companies developing software or software-intensive systems. Besides point estimates, effective project management also requires information about cost-related project risks, e.g., a probability distribution of project costs. One possibility to provide such information is the application of Monte Carlo simulation. However, it is not clear whether other simulation techniques exist that are more accurate or efficient when applied in this context. We investigate this question with CoBRA®, a cost estimation method that applies simulation, i.e., random sampling, for cost estimation. This chapter presents an empirical study, which evaluates selected sampling techniques employed within the CoBRA® method. One result of this study is that the usage of Latin Hypercube sampling can improve average simulation accuracy by 60% and efficiency by 77%. Moreover, analytical solutions are compared with sampling methods, and related work, limitations of thestudy, and future research directions are described. In addition, the chapter presents a comprehensive overview and comparison of existing software effort estimation methods.
Reference:
M. Kläs et al., "The use of simulation techniques for hybrid software cost estimation and risk analysis", in Advances in computers, M. V. Zelkowitz, Ed., London: Academic Press, 2008, pp. 115-174.
Bibtex Entry:
@INCOLLECTION{Klaes2008,
author = {Kläs, Michael and Trendowicz, Adam and Wickenkamp, Axel and Münch,
Jürgen and Kikuchi, Nahomi and Ishigai, Yasushi},
title = {The use of simulation techniques for hybrid software cost estimation
and risk analysis},
booktitle = {Advances in computers},
publisher = {Academic Press},
year = {2008},
editor = {Marvin V. Zelkowitz},
volume = {74},
series = {Software Development},
pages = {115-174},
address = {London},
abstract = {Cost estimation is a crucial field for companies developing software
or software-intensive systems. Besides point estimates, effective
project management also requires information about cost-related project
risks, e.g., a probability distribution of project costs. One possibility
to provide such information is the application of Monte Carlo simulation.
However, it is not clear whether other simulation techniques exist
that are more accurate or efficient when applied in this context.
We investigate this question with CoBRA®, a cost estimation method
that applies simulation, i.e., random sampling, for cost estimation.
This chapter presents an empirical study, which evaluates selected
sampling techniques employed within the CoBRA® method. One result
of this study is that the usage of Latin Hypercube sampling can improve
average simulation accuracy by 60% and efficiency by 77%. Moreover,
analytical solutions are compared with sampling methods, and related
work, limitations of thestudy, and future research directions are
described. In addition, the chapter presents a comprehensive overview
and comparison of existing software effort estimation methods.},
isbn = {978-0-12-374426-5},
issn = {0065-2458},
keywords = {simulation; Monte Carlo method; cost estimation; effort estimation;
COBRA; quantitative analysis}
}