Supporting Many-Objective Software Requirements Decision: An Exploratory Study on the Next Release Problem

Abstract

The decision of which requirements should be satisfied in the next release is crucial to software company. The next release problem, a family of requirements selection decision, aims to maximize profits by satisfying requirements to balance customer profits and development costs. However, due to diverse practical scenarios, solutions to the next release problem have to face many different objectives. In this paper, we propose an exploratory study on the many-objective next release problem with five evolutionary optimization algorithms. The goal of this study is to use the experimental results to assist project managers to make the requirements decision in the scenario of many decision objectives. This study focuses on four research questions, including the effectiveness of optimization, the significance of results, the distribution of metric values, and the correlation between metrics. We design the study to explore five objective functions of the next release problem, including the maximum of customer profits, the minimum of requirements costs, the fairness of requirements selection, etc. The study is conducted on 12 benchmark instances from three real-world projects via evaluating six metrics. Our study suggests that among algorithms in comparison, either eMOEA or IBEA is the best choice for the many-objective next release problem.

Keywords

• Requirements decision
• the many-objective next release problem
• fairness analysis
• search-based requirements engineering