Regionalized sensitivity analysis with respect to multiple outputs - and an application for real-time building space exploration

Building design involves a large number of design parameters and performance indicators. The Monte Carlo method enables the modeler to perform thousands of building performance simulations representing a global design space. To explore such multivariate data (Factor Mapping), the parallel coordinate plot (PCP) is a popular tool, because it is easy to use in “real-time” – even for multiple decision-makers.

However, the PCP becomes unmanageable if it contains many variables, e.g. more than 10–15. Since building simulations typically involve a lot more parameters, we would like to reduce the number of variable inputs (Factor Fixing) while considering their influence towards multiple outputs. Moreover, we would like a method to highlight changes in the PCP, which would allow us to use more variables in the PCP.