DOI: 10.31127/tuje.1527734 ISSN: 2587-1366

A formal and integrated approach to engineering machine learning processes: A method base for project management

Murat Paşa Uysal
Show PDF Cite
Improving project management (PM) capability for machine learning (ML) requires systematic acquisition, structuring, and application of PM knowledge. On the other hand, there are substantial differences between ML-enabled software products (MLESP) and traditional software products (TSP) regarding PM processes. In contemporary tool-centric ML development environments, it is not easy to design and build a method base for PM activities to enable team learning and knowledge management. The studies and industrial applications also indicate that a “one-size-fits-all approach” for PM can be impractical or fall short of the expectations of teams and organizations. The main issue lies in how to elicit, store, and reuse the team’s tacit knowledge related to the methods, processes, tasks, and tools for ML PM. Due to the black box, data-centric, and experimental nature, ML processes can be easily evolved into ad hoc processes. ML teams usually struggle with integrating the ML lifecycle into their software development lifecycle. Therefore, adopting or tailoring a PM method for the specific demands of MLESP emerges as a significant challenge. To address this problem, this study adopted a mixed research approach that harmonized Design Science Research (DSR), PM, Method Engineering (ME) and Process Algebra (PA) activities. An ME framework for PM, a method base, and a novel hybrid PM method formed the main research outputs, which were also expected to meet the requirements of Baskent University Hospital Ankara. The use case-based scenario analysis technique was used for validating the requirements phase of the hybrid PM method. The proposed approach can offer comprehensive, yet pragmatic and adaptable solutions as it blends the strengths of ML, PM, ME, and PA knowledge domains. Additionally, PA played an important role in providing formal and mathematical foundations for the specification and validation of the PM methods and tailoring processes

More from our Archive