Enhanced façade design: A data-driven approach for decision analysis based on past experiences

Abstract

The selection of an optimal building façade system is a challenging process that can be facilitated by using decision-analysis methods. However, current commonly-used decision-analysis tools in civil engineering cannot deal with the interactions among multiple design criteria. The Choquet integral is the only well-known method capable of accounting for such interactions. However, the process of assigning the fuzzy measures (importance weights) for this method is complex, particularly when there is a large number of criteria be considered. This paper proposes two supervised methods to estimate these fuzzy measures. The first method estimates the relative importance weights by using a statistical approach based on Principal Component Analysis, while the second method is elicited from a machine learning algorithm using Neural Networks. These two methods are used in an illustrative example to find the fuzzy measures related to façade design with respect to four criteria; and their merits and limitations are discussed.