Improving the Analytic Hierarchy Process for Requirements Prioritization Using Evolutionary Computing
Juan Carlos B. Somohano-Murrieta, Jorge Octavio Ocharán-Hernández, Ángel J. Sánchez-García, Xavier Limón, María de los Ángeles Arenas-Valdés · Programming and Computer Software · 2021
Abstract Modern software development faces several challenges related to fast technological changes and shorter delivery cycles. This, is reflected in adopting agile approaches that try to lean as much as possible the software development process, from requirements engineering to system deployment. As part of this leaning effort, requirements prioritization is an important activity, as it is usually not possible to implement all requirements at once. Nevertheless, requirements prioritization is a complex activity where several stakeholder’s opinions have to be taken into account hierarchically. There are many techniques to prioritize requirements, being the most accurate and complex the Analytic Hierarchy Process (AHP). AHP is very reliable when prioritizing requirements in the most accurate way due to its mathematical founding, however, this method involves matrix and vectors operations as well as a defined number of pairwise comparisons, which makes it a CPU-intensive method. This paper presents an improved AHP using evolutionary computing techniques to try to optimize requirements prioritization. The proposed approach reduces the complexity of the overall AHP process, and shows to outperform traditional AHP in terms of prioritization accuracy, according to some industry-level requirements data sets extracted from developed software systems.