Evolutionary Robust Optimization for Software Product Line Scoping

Date
2016
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Software product line (SPL) scoping is an important phase when planning for product line adoption. An SPL scope specifies: (1) the extent of the domain supported by the product line, (2) portfolio of products in the product line and (3) list of assets to be developed for reuse across the family of products. An important challenge with SPL scope planning is how estimates may affect the robustness of plans. Scope plans are usually based on estimates about the state of the market and the engineering capabilities of the development team. These estimates are prone to inaccuracy due to the uncertainty in the environment or the measurement accuracy. Therefore, plans based on these estimates may lose optimality or even become infeasible when estimates are inaccurate. To address the above, we propose to include uncertainty as part of the SPL scoping model. Plans developed in consideration of uncertainty would be more robust against possible fluctuations in estimates. In this research, a method to incorporate uncertainty in scoping optimization and its application to generate robust solutions is proposed. We capture uncertainty as part of the formulation and model scoping optimization as a multi-objective problem with profit and stability as fitness functions. More specifically, profit stability and feasibility stability are considered to represent uncertainty concerns. To measure the effectiveness of our proposal and evaluate its validity, a number of different experiments are conducted. Results show that, compared to other scope optimization approaches, both profit stability and feasibility stability are improved while maintaining near-optimal performance for profit objective. Also, generated results consist of solutions with trade-offs between profit and stability, providing the decision maker with enhanced decision support. Multi-objective optimization with stability consideration for SPL scoping provides project managers with a robust and flexible way to address uncertainty in the process of SPL scoping.
Description
Keywords
Computer Science
Citation
Karimpour, R. (2016). Evolutionary Robust Optimization for Software Product Line Scoping (Doctoral thesis, University of Calgary, Calgary, Canada). Retrieved from https://prism.ucalgary.ca. doi:10.11575/PRISM/26004