AutoBDA: Model-Driven Reference Architecture for Automated Big Data Analysis Framework

The formidable growth of the Internet of Things and quality of service requirements accelerate demands for edge-based data-driven services, a major goal of Industry 4.0 and Society 5.0. However, Big Data analysis (BDA) comprehends a diverged stepped process that consumes excessive semantic knowledge and extensive makespan. These limitations hamper the meaningful adoption of BDA and hinder achieving the Industry 4.0 and Society 5.0 goals. Therefore, a solution featuring agile, able to deliver edged-user-friendly automated BDA (AutoBDA), is one of the preferred ways to address the aforementioned concerns. Moreover, BDA is an evolving field in data science. Nevertheless, ad-hoc architectures inherit severe adaptability constraints. Furthermore, solutions that address unique or domain-specific requirements became accustomed to practice; however, they constrain inclusivity. Therefore, we perceived that a holistic modeling approach featuring inclusive and agile facilitates achieving the Industry 4.0 and Society 5.0 goals. Software reference architecture (SRA) is a well-known holistic approach and can alleviate ad-hoc concerns. Automatic service composition (ASC) is preferred to automate diverged-stepped processes. Therefore, we proposed a model-driven holistic SRA for the ASC-based AutoBDA, featuring inclusive and agile up to domain-independent data mining and machine learning requirements. Experiments demonstrate that our proposal efficiently and effectively achieved our objectives.