id |
caadria2023_300 |
authors |
Okhoya, Victor and Bernal, Marcelo |
year |
2023 |
title |
Variability in Machine Learning for Multi-Criteria Performance Analysis |
source |
Immanuel Koh, Dagmar Reinhardt, Mohammed Makki, Mona Khakhar, Nic Bao (eds.), HUMAN-CENTRIC - Proceedings of the 28th CAADRIA Conference, Ahmedabad, 18-24 March 2023, pp. 149–158 |
doi |
https://doi.org/10.52842/conf.caadria.2023.1.149
|
summary |
Parametric analysis is emerging as an important approach to building performance evaluation in architectural practice. Since architectural performance has many competing metrics multi-criteria analysis is required to deal effectively with the complexity. However, multi-criteria parametric analysis involves large design spaces that are expensive to compute. Machine learning is emerging as an important design space reduction method for multi-criteria analysis. However, there are many types of machine learning algorithms and architects can benefit from understanding which algorithms perform well on which tasks. Using a mid-rise commercial residential tower project this paper investigates three common machine learning algorithms for performance against three common performance metrics. The algorithms are multi-layer perceptrons, support vector machines, and random forests, while the metrics are site energy, illuminance, and a value function that combines them both. In addition, we seek to understand what factors are most impactful in improving algorithm performance. We investigate four impact factors namely sample size, sensitivity analysis, feature selection, and hyperparameters. We find that multi-layer perceptrons perform best for all three performance metrics. We also find that hyperparameter tuning is the most impactful factor affecting multi-layer perceptron performance. |
keywords |
parametric analysis, machine learning, design space |
series |
CAADRIA |
email |
|
full text |
file.pdf (1,817,084 bytes) |
references |
Content-type: text/html
Access Temporarily Restricted
Access Temporarily Restricted
Too many requests detected. Please wait 60 seconds or verify that you are a human.
If you are a human user and need immediate access, you can click the button below to continue:
If you continue to experience issues, please open a ticket at
papers.cumincad.org/helpdesk
|
last changed |
2023/06/15 23:14 |
|