Improving Software Effort Estimation through a Hybrid Approach of Metaheuristic Algorithms in Analogy-based Method

Project management in software development is one of the most crucial activities as it encompasses the entire software development process from start to finish. Estimating the effort required for software projects is a significant challenge in project management. Managing software projects and consequently estimating their effort for more efficient and impactful management of such projects is necessary and unavoidable. Analogy-based estimation in software effort estimation involves comparing new projects to completed ones. However, this method can be ineffective due to variations in feature importance and dependencies. To address this, weights are assigned to features using optimization techniques like meta-heuristic algorithms. Yet, these algorithms may get stuck in local optima, yielding nonoptimal results. An approach to estimate software effort is proposed in this study. It aims to find global optimal feature weights by combining particle swarm and genetics metaheuristic algorithms. This hybrid approach leverages particle motion and composition to enhance solution generation, increasing the likelihood of finding the global optimum and overcoming local optima issues. The algorithm calculates feature weights for project estimation using analogy-based methods. The proposed approach was tested and assessed using two datasets, namely Maxwell and Desharnais. The experimental results indicated an enhancement in the evaluation criteria, including MMRE, MdMRE, and PRED, compared to similar research works.