Requirements Elicitation using “Internet of Things” (IoT) Data: A conceptual meta-model for the elicitation of requirements from IoT data.

Sammanfattning

The continuous digital transformation led to the generation of vast amounts of data characterized by their volume, variety, and velocity. While this data brings out the potential for eliciting system requirements and, therefore, for developing systems, it challenges traditional requirement elicitation methods. Conventional methods cannot efficiently consider and manage dynamic, complex, and unstructured data generated by humans or machines. Consequently, there have been many efforts in the research community aiming at data-driven requirements elicitation. Most of these efforts focus on human-generated rather than machine-generated data. While some efforts focus on the latter, they are mostly less generic and narrow. To fill this research gap, this thesis proposes a meta-model that conceptualizes the elicitation of requirements from Internet of Things (IoT) data. This study explores collecting data from sensors, processing the data, including behavioral analysis, and mapping the analyzed and aggregated data to potential requirements. Within the IoT context, this is particularly relevant because the IoT is anticipated to generate an increasing amount of data as their sensor potentials are expected to be gradually more exploited. Following the Design Science Research (DSR) framework by Johannesson & Perjons,  this study explores the challenges involved in the development process of an artifact, developing a structured meta- model, and conducting an expert evaluation. The study follows the following five steps of the mentioned DSR framework: Explicate the Problem, Define Requirements, Design and Develop the Artifact, Demonstrate the Artifact, and Evaluate the Artifact. Within each of these steps, different research strategies were employed. Systematic literature reviews have been conducted to explicate the problem and define the requirements. The design and development of the meta-model relied more on the previously conducted literature review than on other research strategies. Further, brainstorming and sketching sessions and discussion with another researcher were employed for the design and development step. For the demonstration and evaluation of the meta-model action research has been applied. More specifically, a walkthrough demonstrated the artifact, examining the logical feasibility, and expert evaluation was conducted through a questionnaire to assess the artifact. The study's findings demonstrate the meta-model's completeness, correctness, applicability, and scalability to various IoT applications. Further, it indicates that the meta-model provides a generic and high-level concept for eliciting requirements from IoT data that can be applied in the next steps of future research to an actual use case scenario, enhancing its validation.