Statement of the problem. Currently, there is an obvious need to develop methods and approaches that allow analytical models to effectively cope with the uncertainty and errors of the source data, which is key to the successful implementation of machine learning in construction projects. Results. The article pays special attention to the reliability of analytical models, which require high accuracy to ensure the safety and durability of structures. The influence of the quality of the initial data, including the representativeness and balance of the sample, is analyzed, and the problems of adapting analytical models are also discussed. The importance of integrating knowledge about the work of the process occurring in structures for training (building) analytical models is noted, which can increase their adequacy and accuracy. Conclusions. The main directions for testing machine learning systems — evaluating the adequacy of the obtained analytical models for construction tasks — are formulated and justified.
Redina et al. (Wed,) studied this question.