A framework of developing machine learning models for facility life-cycle cost analysis

Xinghua Gao and Pardis Pishdad-Bozorgi
Building Research & Information
Machine learning techniques have been used for predicting facility-related costs but there is a lack of research on developing machine learning models for the complete life-cycle cost (LCC) analysis of facilities. This research aims to systematically investigate the feasibility of forecasting facilities' LCC by implementing machine learning on historical data. The authors propose a comprehensive and generalizable framework for developing facility LCC analysis machine learning models. This framework specifies the data requirements, methods, and expected results in each step of the model development process. First, a literature review and a questionnaire survey were conducted to determine the independent variables affecting facility LCC and to identify the potential data sources. The process of using raw data to derive LCC components is then discussed. Finally, a proof-of-concept case study was conducted on a university campus to demonstrate the application of the proposed framework. This research concludes that current building systems already contain the data for LCC analysis and that the proposed framework is effective in facility LCC prediction.