Process-oriented guidelines for systematic improvement of supervised learning research in construction engineering

A limited assessment of the development process and various stages of machine learning (ML) based solutions for construction engineering (CE) problems are available in the literature. After critically reviewing 467 published articles from 2017 to 2022, sixteen areas of improvement at different stages of ML projects' development cycle are identified and discussed with examples from past studies. These areas of improvement are categorized along different stages of the development cycle of an ML-based engineering solution: 1) Justification of applying ML models, 2) employing the proper data collection strategy, 3) using accurate data pre-processing techniques, 4) selecting and training a model correctly, and 5) performance validation. Researchers and practitioners can utilize the results of this review to enhance the reliability of future trained ML models and broaden the applicability of research-oriented ML-based CE studies. Our findings can assist in bridging the gap between the research and practice of ML-based CE-related studies.

Automated summary

A limited assessment of the development process and various stages of machine learning (ML) based solutions for construction engineering (CE) problems are available in the literature. Sixteen areas of improvement at different stages of ML projects' development cycle are identified and discussed with examples from past studies. These areas of improvement include justifying the application of ML models, employing proper data collection strategies, using accurate data pre-processing techniques, selecting and training models correctly, and performance validation. Researchers and practitioners can utilize these findings to enhance the reliability and applicability of ML-based CE studies, bridging the gap between research and practice.