RAIDAD: A model-driven framework for automated and agile development of IoT data analysis software

Abstract

Context: Nowadays, developing data analysis software for the IoT domain faces challenges such as complexity, repetitive tasks, and developers’ lack of domain knowledge. To address these issues, methodologies like CRISP-DM have been introduced, providing structured guidance for data analysis. Objectives: Despite the availability of structured methodologies, building data analysis pipelines still involves managing complexity and redundancy. Model-driven approaches have been proposed to tackle these challenges but often fail to address all stages of the data analysis workflow and the interdependencies between stages and datasets comprehensively. This research introduces RAIDAD, a model-driven framework that addresses these gaps by covering all phases of the CRISP-DM methodology. Methods: RAIDAD includes a domain-specific modeling language for IoT data analysis, a graphical modeling editor, a code generation transformation engine, and a data model assistant for seamless model-data integration. These components are delivered as an Eclipse plugin. Results: The evaluation of RAIDAD is two-fold. First, a comparative operational evaluation with RapidMiner and ML-Quadrat shows RAIDAD achieves a 9.6% improvement in usability and productivity over RapidMiner and a 23% improvement over ML-Quadrat. Second, RAIDAD is compared to a general-purpose programming language, demonstrating its superiority in reducing effort and production time for IoT data analysis software. Conclusion: This comprehensive framework ensures an efficient and organized approach to data analysis, addressing key challenges in the IoT domain. Future research will focus on expanding RAIDAD's support for a wider range of data analysis and machine learning algorithms, enhancing automation capabilities, and incorporating continuous user feedback to ensure the framework evolves in line with emerging needs. © 2025 Elsevier B.V.