[Bldg-sim] ISO Precedent paper in multi-objective optimization: statistical comparison of two pareto sets

[Thomas Wortmann]

Dear Holly,

I'm not aware of a best practice for this.

One idea would be to evaluate the non-dominated set A according to the criteria for B, and vice-versa.

You could then compare the resulting fronts with standard metrics such as the hypervolume.

Hope that helps!

Best,
Thomas
____________________________

Thomas Wortmann, Ph.D. M.Sc. M.Arch



Tenure-Track Professor, Chair for Computing in Architecture

Faculty, International Max Planck Research School for Intelligent Systems


ICD – Institute for Computational Design and Construction

Faculty for Architecture and Urban Planning

University of Stuttgart
Keplerstr. 11
70174 Stuttgart

Germany


Tel +49 (0) 711 685 81933
________________________________
From: Bldg-sim <bldg-sim-bounces at lists.onebuilding.org> on behalf of Samuelson, Holly Wasilowski via Bldg-sim <bldg-sim at lists.onebuilding.org>
Sent: Wednesday, July 14, 2021 4:48:00 PM
To: bldg-sim at lists.onebuilding.org <bldg-sim at lists.onebuilding.org>
Cc: Salimi, Shide <ssalimi at gsd.harvard.edu>
Subject: [Bldg-sim] ISO Precedent paper in multi-objective optimization: statistical comparison of two pareto sets

External Email – Do not click on any links or open any attachments from an unknown sender.



Dear Building Sim,
Can any of you recommend a precedent paper for us?  In short, we want to compare two sets of optimal solutions in two different multi-objective optimizations, to determine whether the characteristics of the two sets are significantly different.

Longer version:
We set up a multi-objective analysis and found a set of optimal solutions (buildings whose simulation results lie on the pareto front).  Then we did this again for slightly different objectives, i.e. different comfort criteria.  We now have two sets of "optimal" buildings. They have different characteristics, like insulation levels and window-to-wall ratios.  We want to determine if the differences between the two sets are significant. We have ideas of how to analyze this, but it would be great to see how others have done this in the past. This precedent research could be from any field using optimization, not just building science.

If you have any leads for us, we'd appreciate it.
Thanks!
Holly

Holly W. Samuelson | DDes | Registered Architect
Associate Professor | Harvard Graduate School of Design
Co-Head MDes Energy and Environment Program
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.onebuilding.org/pipermail/bldg-sim-onebuilding.org/attachments/20210715/c0e892b0/attachment-0002.html>