Modeling of Software Development Effort A Simulated Annealing Approach

Abstract — To accurately model the inter-relationship between the software project effort/cost and the effort/cost drivers various modeling studies have been conducted. These mainly include empirical and soft computing and expert analogy based techniques, primarily aimed at accurate software estimation and efficient management of software projects. This paper presents a Simulated Annealing (SA) based optimization approach using Matlab to minimize the variance in software development (SD) effort modeling of NASA SD projects. A comparison of results based on SA approach with non-linear regression and other available models is also presented proving the estimation utility of the SA approach. Keywords-software development, effort estimation, simulated annealing, variance, regression I