Answers are grounded in this paper and cite their sources.
Evaluates a novel hyper-heuristic in multi-objective software cost estimation, indicating improved performance.
Software companies estimate different costs involving any assignment, known as software cost estimation (SCE). However, their goal is manifold; most of the studies on SCE follow minimization of the mean magnitude of relative error (MMRE) of estimation. Here, a hyper-heuristic for a multi-objective continuous optimization problem (MOCOP) is designed and used for multi-objective SCE (MOSCE). The study uses Constructive Cost Model and several data sets, including NASA-93. Three basic heuristics, Artificial Bee Colony, Genetic Algorithm, and Grey Wolf Optimizer, are modified using the non-dominated sorting property of NSGA-II for solving the MOCOP and are named NSMOABC, NSMOGA, and NSMOGWO, respectively. These are coordinated using Q-learning to develop a hyperheuristic, named Multi-objective Metaheuristic using Q-learning (MOMHQ), for the MOCOP. It is observed that in every independent execution, MOMHQ consistently reaches the true Pareto optimal front for every considered test instance. The diversity of objective vectors on the output Pareto front suggests that MOMHQ effectively explores and covers the entire solution space. The performance of MOMHQ is statistically compared with a set of state-of-the-art algorithms, and its superiority is established. It is observed that the performance of MOMHQ to MOSCE is better compared to existing studies.
No takes yet. Share an insight, caveat, or question.
Bera et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: