28-09-2012, 04:21 PM
CONSTRUCTIVE COST ESTIMATION MODEL(COCOMO) BASED EVOLUTIONARY COMPUTATIONAL METHODS
Abstract
Software development effort prediction is one of the most critical activities in managing software projects. Algorithmic effort prediction models, which have dominated the software engineering community, are limited by their inability to cope with uncertainties and imprecision surrounding software projects early in the development life cycle. More recently, attention has turned to a variety of machine learning methods, and soft computing in particular to predict software development effort. There are evidences that soft computing has been able to address some of the problems associated with previous models. However, there is no common ground for assessing and comparing these soft computing based prediction techniques.
In this study, a multivariate interpolation model was developed to estimate the effort component and different cost drivers of the software projects. A COCOMO based equation was used to represent the effort function. The data set that was used consists of two independent variables, first is Lines of Code(LOC) and second is Methodology (ME) and one dependent variable Effort (CE). Data set is taken from NASA projects. It can be shown that Evolutionary Computational Methods can be used to estimate the optimal parameters of the effort components of software projects. The aim of this study is to analyze soft computing techniques in the existing models and to provide in depth review of software and project estimation techniques existing in industry and literature based on the different test datasets along with their strength and weaknesses[1].
Related studies
A great deal of interest has appeared in the literature on the application of guided search techniques to different areas. The paper [2] compares four search mechanisms, tabu search, simulated annealing, genetic algorithms and branch and bound, in structural engineering optimization problems. According to this study, tabu search, simulated annealing and genetic algorithms work well and produce an acceptable sub-optimal solution within a reasonable amount of time. Branch and bound is not a viable search method for structures of any reasonable size.
Tabu search arrives at the solution quicker than both simulated annealing and genetic algorithms.