Ensembling Harmony Search Algorithm with case-based reasoning for software development effort estimation

Estimating software development effort is challenging in ensuring timely completion of projects and managing resources in software development companies. Inaccurate estimation of development efforts leads to significant financial losses and delays in the software project’s completion. Due to dynamic requirements, technological evolution, inadequate historical data, human factors, and project complexity, the developed models cannot achieve satisfactory accuracy. Case-based reasoning (CBR) is a technique that uses data from previous projects to estimate the effort of the new project by identifying and adapting solutions that were successful in similar contexts. However, the effectiveness of CBR depends on tuning its multiple parameters, such as how past similar projects are retrieved, reused, adapted, and retained. In this paper, the Harmony Search Algorithm (HSA) is used to identify the best combination of traditional CBR parameters (feature selection, similarity measures, and the k-value count of closest neighbors, feature weighting) to accurately estimate the development effort. This paper uses the HSA to optimize the parameters for CBR, enhancing the accuracy of the estimation. The proposed CBR-HSA approach is validated using thirteen public datasets from the PROMISE repository, NASA, SEERA, and a subset of the ISBSG dataset. It is evaluated using six reliable evaluation metrics. The results obtained are promising, particularly in accuracy, statistical significance, and computational time compared to some existing models.