Victor Okhoya will present the defense for his thesis, "Machine Learning for Parametric Analysis in Architectural Practice," to obtain the Doctor of Professional Practice (DPP/DDES) on Tuesday 19 November.
Title: “Machine Learning for Parametric Analysis in Architectural Practice”
By Victor Okhoya, Doctoral Candidate
Date: Tuesday, 19 November 2019
Location: MMCH A11
PhD Advisory Committee
Ramesh Krishnamurti (Chair; CMU School of Architecture)
John Haymaker (Perkins and Will)
Arati Singh (CMU Computer Science)
Daniel Cardoo Llach (CMU School of Architecture)
Parametric analysis is an emergent, data-driven approach to building performance analysis in architectural design. It has the benefits of enabling the optimization of design spaces, reducing uncertainty in design decision making, and establishing the most sensitive input parameters of the design space. However, it also faces data analytical challenges in practical application. These challenges include the size of design spaces, the speed and accuracy of simulation, and the presence of complex design conditions. This study proposes to use machine learning algorithms as a design space reduction strategy for overcoming these data analytical challenges and thereby facilitating the implementation of parametric analysis in architectural practice. The study asks if machine learning algorithms can be effective at design space reduction for multi-discipline parametric analysis; which algorithms, in particular, are the most effective? Are machine learning algorithms robust under complex design conditions? Further, which impact factor among sample size, sensitivity analysis, feature selection, and hyperparameter tuning are most effective in improving algorithm performance? The study finds that machine learning algorithms can indeed be effective for multi-discipline parametric analysis. In particular, artificial neural networks are the most effective algorithm. Likewise, hyperparameter tuning is the most influential impact factor affecting algorithm performance. Further, machine learning algorithms are robust under complex design conditions.