One of the reasons for unsuccessful machine learning (ML) applications can be regarded as unsatisfactory requirements engineering (RE). processes. The data-driven and black box nature of ML is a great obstacle for RE, design, and development processes. RE approaches, such as Goal-Oriented RE or Value-Based RE, focus on the organizational context, and they adopt a business value perspective. Viewpoint-Based RE methods that use software engineering-specific tools and techniques are useful, but, usually difficult to understand for non-technical stakeholders. Therefore, Scenario-Based RE (SBRE) methods would be beneficial and can form a mutual understanding. However, they have engineering limitations, and can easily lead to ambiguous requirements specifications if they rely only on text, and natural language descriptions, or graphical representations. Consequently, Use Case Maps (UCM) can serve as a bridge between requirements and design processes, and they can provide a visual, scenario-based, behavioral, structural, and causal path-based framework for RE. However, the review of literature cannot provide the studies exploring UCM as an SBRE tool for ML applications. A scenario-based approach is adopted for the RE processes of ML applications. Eclipse Modeling Tools IDE and jUCMNav v7.0.0 plug-in are used for the design and development of UCM. The main contribution of this study is the introduction of both scenario-based approach and UCM as a RE tool for ML
Anahtar Kelimeler: Machine Learning, Requirements Engineering, Scenario-Based Requirement Engineering, Use Case Map
|
