Pages: 9-20

Abstract

Vast studies have recently been conducted on increasing the accuracy of effort estimation, one of the most important one of which is COCOMO II. Another challenge in software development management is effort distribution among the phases of software development. In the proposed method, in order to estimate and distribute effort to develop software product, hybrid methods by the use of several effort estimation methods can increase the accuracy of estimation. In this research, by presenting a hybrid model of effort estimation in software development, attempts have been made to reduce deviations in effort distribution at each phase in software development and finally to increase the accuracy of effort estimation. The presented effort estimation method is based on a combination of fuzzy expert system and experimental effort estimation model COCOMO II. Having conducted the evaluations using NASA data set, level of increase of effort estimation in the presented method was investigated using MARE and VARE metrics. Data analysis indicates a reduction in the level of evaluation metrics and consequently an increase in the accuracy of effort estimation using the proposed method.