An automatic methodology for the quality enhancement of requirements using genetic algorithms

Adanza Dopazo, Daniel; Moreno Pelayo, Valent�n; G�nova Fuster, Gonzalo

Abstract

Context: The set of requirements for any project offers common ground where the client and the company agree on the most important features and limitations of the project. Having a set of requirements of the highest possible quality is of enormous importance; benefits include improving project quality, understanding client needs better, reducing costs, and predicting project schedules and results with greater accuracy. Objective: This paper's primary goal is to create a methodology that can provide effective and efficient solutions for modifying poor requirements integrated into a full-fledged system, extracting the main features of each requirement, assessing their quality at an expert level, and, finally, enhancing the quality of the requirements. Method: In the first step, a machine learning algorithm is implemented to classify requirements based on quality and identify those that are the likeliest to be problematic. In the second step, the genetic algorithm generated solutions to enhance the quality of the requirements identified as inferior. Results: The results of the genetic algorithm are compared with the theoretically optimal solution. The paper demonstrates the significant flexibility of genetic algorithms, which create a wide variety of solutions and can adapt to any type of classifier. From the initial dataset of requirements, the genetic algorithm finds the optimal solution in 85% of cases after 10 iterations and achieves 59.8% success after only one iteration. Conclusions: Genetic algorithms are promising tools for requirements engineering by delivering benefits such as saving costs, automating tasks, and providing more solid and efficient planning in any project through the generation of new solutions.

• Journal Article Type: Article
• Acceptance Date: Jul 23, 2021
• Online Publication Date: Jul 24, 2021
• Publication Date: Dec 1, 2021
• Deposit Date: Aug 22, 2022
• Journal: Information and Software Technology
• Print ISSN: 0950-5849
• Publisher: Elsevier
• Peer Reviewed: Peer Reviewed
• Volume: 140
• Pages: 106696
• DOI: https://doi.org/10.1016/j.infsof.2021.106696
• Keywords: Computer Science Applications; Information Systems; Software, Requirements engineering Genetic algorithm Requirements quality
• Public URL: https://uwe-repository.worktribe.com/output/9852459
• Publisher URL: https://www.sciencedirect.com/science/article/pii/S0950584921001518?via%3Dihub