Multi-Objective Optimization for Software Testing Effort Estimation.
Abstract
Software Testing Effort (STE), which contributes about 25-40% of the total development effort, plays a significant role in software development. In addressing the issues faced by companies in finding relevant datasets for STE estimation modeling prior to development, cross-company modeling could be leveraged. The study aims at assessing the effectiveness of cross-company (CC) and within-company (WC) projects in STE estimation. A robust multi-objective Mixed-Integer Linear Programming (MILP) optimization framework for the selection of CC and WC projects was constructed and estimation of STE was done using Deep Neural Networks. Results from our study indicate that the application of the MILP framework yielded similar results for both WC and CC modeling. The modeling framework will serve as a foundation to assist in STE estimation prior to the development of new a software project. Keywords- Software Testing Effort; Cross-Company; WithinCompany; Optimization; Deep Neural Networks I. INTRODUCTION Software Testing Effort (STE) is one of the critical phases in software development as it is estimated to contribute about 2540% [1] of the total development cost. STE is the effort in relation to duration, resources and source code for testing the software before deployment to the client or market [1]. The acquisition of relevant data for effort estimation has been a focus for prior work [2] concerning the effective use of withincompany (WC) and cross-company (CC) datasets for predictive modeling. Researchers and practitioners who lack the relevant local data rely on data imported from other repositories or companies. The challenge is that this approach might not always be useful for WC estimation. In order for WC estimation to benefit from CC datasets, the selection of an optimal "mix" of project datasets and features is required for efficient estimation. The WC and CC modeling have been considered in the domains of software development effort estimation and defect prediction [2][4]. To the best of our knowledge, this is the first study to employ the use of CC projects in the estimation of STE. A study by Burak et al. [4] indicates that models that used CC datasets f