Breaking into the black box of ESG in the building sector: a machine learning approach

Principal Investigators 

Prof. Dr. Bing Zhu
Professorship for Real Estate Development, Department of Civil, Geo and Environmental Engineering, TUM School of Engineering and Design

Prof. Dr. Debarghya Ghoshdastidar
Professorship for Theoretical Foundations of Artificial Intelligence, Department of Informatics, TUM School of Computation, Information and Technology

Project summary

This project aims at providing AI solutions to the following problems in the Environmental, Social, and Governance (ESG) assessment in the building sector, namely the cost problem, the arbitrariness problem, the missing data problem, and the opaqueness problem. As demand for ESG grows, investors are set to start demanding more accurate and timely responses to ESG issues. However, due to the lack of comparability of ESG data, metrics, and approaches, the ESG rating is still a black box, which becomes a significant bottleneck for sustainable development in the building sector.

To solve this problem, this project will improve the ESG assessment by filling in missing data and optimizing the weighting scheme. Methodologically, this project will apply advanced machine learning methods, including matrix completion and graph-based semi-supervised learning to complete the data gap. From a machine learning perspective, predicting ESG assessment is an example of a small data problem, where the available training data is too few to apply powerful deep learning tools. Hence, this project will develop new algorithms for small data by using concepts from the statistical learning theory. Finally, with the completed data and improved weighting scheme, this project will also fill in the literature gap on assessing the real impact of ESG KPIs on real estate valuation and investment performance.