Transfer learning for smart construction: Advances and future directions

Abstract

Transfer learning has emerged as a powerful tool and rapidly advanced numerous fields with cutting-edge technologies. This paper provides a comprehensive review of transfer learning applications in smart construction, analyzing its utilization to enrich the construction industry's knowledge. A systematic analysis of 366 publications from 2015 to 2024 highlights the growth and importance of transfer learning in the field. This review establishes a foundational framework by exploring key questions: “Why transfer learning”, “What to transfer”, “How to transfer”, and “When to transfer”. The findings reveal that transfer learning is predominantly applied in seven key construction domains, but it faces four major challenges: “transfer strategy”, “interpretability, security and privacy”, “modality transfer”, and “cross-domain adaptability”. Corresponding future research directions are proposed to address these challenges. This paper serves as a crucial reference point for researchers, practitioners, and stakeholders aiming to harness the transformative potential of transfer learning in the construction industry.