An approach for requirements prioritization based on tensor decomposition

A potential mathematical framework for machine learning is multi-linear algebra of the higher-order tensor that can reveal the relationships among multiple factors underlying the observations. Prioritizing the requirements of a project facilitates the process of requirements engineering and involves multifactors. Due to existing time constraints and budget related to projects, by prioritizing the requirements in an appropriate order we can select and apply them more accurately and this causes to increase the quality of software and customers’ satisfaction. In order to prioritize the requirements, there are many approaches that consider different parameters and different view point in their prioritization process. But as far as we know none of them considers the simultaneous effect among entities, namely functional requirements, non-functional requirements and stakeholders in their prioritization process. In this paper, we decided to consider the simultaneous effect among functional, non-functional requirements and stakeholders that have different preferences on requirements by modeling a three-order tensor. Then by applying multi-way analysis, we will obtain appropriate ordered lists of requirements. To evaluate our approach, a controlled experiment has been provided that compares the proposed approach with the state-of-the-art-based approach, analytic hierarchy process (AHP). The results show that our proposed approach outperforms AHP in terms of actual time consumption and ease of use while preserving the quality of the results obtained by our proposed approach.