An adaptive AI-MCDM framework for sustainability assessment in construction projects

The construction sector is a significant producer of greenhouse gas emissions and depletors of natural resources, but its application of sustainability assessment is fragmented and separate from the use of computerized tools for managing building and construction projects. In this study, an adaptive AI-enhanced multiple criteria decision making (MCDM) framework that integrates expert opinions and recalibrates a neural network using a multilayer perceptron (MLP) has been developed. Also, the framework includes explainable AI via SHAP analysis and benchmarking of results to international standards (LEED, BREEAM, and DGNB). Three construction projects in Albania were selected to test the framework; Project A and Project B obtained sustainability scores of 0.825 and 0.813, respectively, and both exceeded the lowest global threshold. However, Project C achieved a score of 0.447 and indicated structural issues in achieving sustainability. Results of SHAP analysis showed that the two most important factors influencing sustainability were the reduction in CO2 emissions (up to 38% of total environmental weight post-AI recalculation) and cost efficiency. Compared to static weights, the AI-enhanced model decreased deviations from international standards by approximately 15%. A sensitivity analysis and a series of Monte Carlo simulations provided evidence of the robustness of the methodology, where changes to project ranking did not exceed 3%. Thus, this method will assist construction managers and policy makers to develop sustainable assessment methods that can be integrated into digital applications.