Machine learning (ML), has become very popular, and it has gained much attention in academia and industry. Besides successful stories, industrial applications and research studies also report plenty of failures, unsatisfactory results, and a great many ML projects failing to meet unrealistic and feasible expectations. One of the important factors is the deficiency in requirements engineering (RE), specifically, regarding organizational contexts, and well-conducted Goal-Oriented Requirements Engineering (GORE) processes. However, the review of literature cannot provide sufficient work and present the required guidelines for GORE of ML applications. In this study, we adopt a GORE approach and extend it for the RE processes of ML applications. Goals can be expressed at different levels of abstraction and they are useful for elicitation, analyzing and identifying alternatives, and conflict resolutions related to RE. The GORE approach focuses on the strategic context of ML requirements and helps elaborate the requirements that support the organizational goals. Therefore, GORE’s main concerns are: how an ML application would achieve the organizational goals, how to operationalize these goals into services or products, and how to assign tasks and responsibilities. Goal-Oriented Requirement Language (GRL) is used for both modeling and reasoning. The contributions of this study are two-fold: (1st) drawing the practitioners’ and researchers’ attention to the problems of GORE for ML, (2nd) proposing an extended version of the GORE model. Consequently, this study can be viewed as an initial step towards the knowledge domain of RE for ML
Anahtar Kelimeler: Machine Learning, Requirements Engineering, Goal-Oriented Requirements Engineering
|
