Uncertainty-permeable Analytics for the Next Release Problem through Fuzzy Mathematical Programming and Possibility Theory

Software development processes pose a number of challenges in terms of the effect of uncertainty on them. This un-certainty has led much of the process to be made more agile, using iterative-incremental lifecycle models, dividing the work into a sequence of releases. In each release, usually of fixed duration, it must be decided which requirements will be implemented, a problem usually referred to as the next release problem. However, there is uncertainty about the effort that needs to be invested in each requirement, and errors in the estimation or the occurrence of certain events can quickly cause the original plan (i.e., the subset of requirements to be implemented in next release) to be adapted to these changes. This paper presents a novel fuzzy next release model that considers the penetration and exploitation of the uncertainty related to the effort needed to implement each requirement. The solution of such model is a consonant family of plans in the form of a possibility distribution over subsets of requirements, which allows determining which requirements to add or postpone as the iteration progresses and changes arise. Such information constitutes a flexible prescriptive analytic that supports decision-making. In addition, a computational empirical study is performed on public instances using state-of-the-art Integer Linear Programming exact solvers to assess whether the computational time needed to solve such a model is adequate for the referred context. The results are promising and pose challenges that encourage further research.