An adaptive multi-objective differential evolution algorithm to fine-tune the constructive cost model constants for software effort estimation

Abstract

The primary goals of software effort and cost estimation are to determine project cost, shorten development time, and provide a reliable prediction. This issue is also classified as a multi-objective optimization problem. A new adaptation-based multi-objective differential evolution algorithm, enhanced with Generative Artificial Intelligence (AI) techniques, solves these problems by tuning the parameters. This paper includes new mutation techniques with a Pareto-based differential evolution algorithm, leveraging Generative AI to increase candidate solution diversity. The new mutation operator provides more bandwidth, which helps the differential evolution algorithm to avoid local optimality problems. This study applies a non-dominated sorting technique to lower the computational complexity involved in Pareto dominance. The article further examines the software cost estimation problem by fine-tuning the parameters for a multi-objective constructive cost model using Generative AI to predict software costs accurately. The software cost estimation problem aims to minimize prediction errors while maximizing accuracy, ultimately reducing overall project costs. Compared to existing modern versions of multi-objective evolutionary optimization algorithms across all objective problems, the proposed approach, which integrates Generative AI, shows improved prediction and lower error rates on the National Aeronautics and Space Administration (NASA)-93 and Constructive Cost Model (COCOMO)-81 datasets.