Improving Design Preference Prediction Accuracy Using Feature Learning

Alex Burnap, Yanxin Pan, Ye Liu, Yi Ren, Honglak Lee, Richard Gonzalez, Panos Y. Papalambros

Research output: Contribution to journalArticle

14 Citations (Scopus)

Abstract

Quantitative preference models are used to predict customer choices among design alternatives by collecting prior purchase data or survey answers. This paper examines how to improve the prediction accuracy of such models without collecting more data or changing the model. We propose to use features as an intermediary between the original customer-linked design variables and the preference model, transforming the original variables into a feature representation that captures the underlying design preference task more effectively. We apply this idea to automobile purchase decisions using three feature learning methods (principal component analysis (PCA), low rank and sparse matrix decomposition (LSD), and exponential sparse restricted Boltzmann machine (RBM)) and show that the use of features offers improvement in prediction accuracy using over 1 million real passenger vehicle purchase data. We then show that the interpretation and visualization of these feature representations may be used to help augment data-driven design decisions.

Original languageEnglish (US)
Article number071404
JournalJournal of Mechanical Design, Transactions Of the ASME
Volume138
Issue number7
DOIs
StatePublished - Jul 1 2016

Fingerprint

Principal component analysis
Automobiles
Visualization
Decomposition

ASJC Scopus subject areas

  • Mechanical Engineering
  • Mechanics of Materials
  • Computer Graphics and Computer-Aided Design
  • Computer Science Applications

Cite this

Improving Design Preference Prediction Accuracy Using Feature Learning. / Burnap, Alex; Pan, Yanxin; Liu, Ye; Ren, Yi; Lee, Honglak; Gonzalez, Richard; Papalambros, Panos Y.

In: Journal of Mechanical Design, Transactions Of the ASME, Vol. 138, No. 7, 071404, 01.07.2016.

Research output: Contribution to journalArticle

Burnap, Alex ; Pan, Yanxin ; Liu, Ye ; Ren, Yi ; Lee, Honglak ; Gonzalez, Richard ; Papalambros, Panos Y. / Improving Design Preference Prediction Accuracy Using Feature Learning. In: Journal of Mechanical Design, Transactions Of the ASME. 2016 ; Vol. 138, No. 7.
@article{eb898d02b64640269390f61781aa99a5,
title = "Improving Design Preference Prediction Accuracy Using Feature Learning",
abstract = "Quantitative preference models are used to predict customer choices among design alternatives by collecting prior purchase data or survey answers. This paper examines how to improve the prediction accuracy of such models without collecting more data or changing the model. We propose to use features as an intermediary between the original customer-linked design variables and the preference model, transforming the original variables into a feature representation that captures the underlying design preference task more effectively. We apply this idea to automobile purchase decisions using three feature learning methods (principal component analysis (PCA), low rank and sparse matrix decomposition (LSD), and exponential sparse restricted Boltzmann machine (RBM)) and show that the use of features offers improvement in prediction accuracy using over 1 million real passenger vehicle purchase data. We then show that the interpretation and visualization of these feature representations may be used to help augment data-driven design decisions.",
author = "Alex Burnap and Yanxin Pan and Ye Liu and Yi Ren and Honglak Lee and Richard Gonzalez and Papalambros, {Panos Y.}",
year = "2016",
month = "7",
day = "1",
doi = "10.1115/1.4033427",
language = "English (US)",
volume = "138",
journal = "Journal of Mechanical Design - Transactions of the ASME",
issn = "1050-0472",
publisher = "American Society of Mechanical Engineers(ASME)",
number = "7",

}

TY - JOUR

T1 - Improving Design Preference Prediction Accuracy Using Feature Learning

AU - Burnap, Alex

AU - Pan, Yanxin

AU - Liu, Ye

AU - Ren, Yi

AU - Lee, Honglak

AU - Gonzalez, Richard

AU - Papalambros, Panos Y.

PY - 2016/7/1

Y1 - 2016/7/1

N2 - Quantitative preference models are used to predict customer choices among design alternatives by collecting prior purchase data or survey answers. This paper examines how to improve the prediction accuracy of such models without collecting more data or changing the model. We propose to use features as an intermediary between the original customer-linked design variables and the preference model, transforming the original variables into a feature representation that captures the underlying design preference task more effectively. We apply this idea to automobile purchase decisions using three feature learning methods (principal component analysis (PCA), low rank and sparse matrix decomposition (LSD), and exponential sparse restricted Boltzmann machine (RBM)) and show that the use of features offers improvement in prediction accuracy using over 1 million real passenger vehicle purchase data. We then show that the interpretation and visualization of these feature representations may be used to help augment data-driven design decisions.

AB - Quantitative preference models are used to predict customer choices among design alternatives by collecting prior purchase data or survey answers. This paper examines how to improve the prediction accuracy of such models without collecting more data or changing the model. We propose to use features as an intermediary between the original customer-linked design variables and the preference model, transforming the original variables into a feature representation that captures the underlying design preference task more effectively. We apply this idea to automobile purchase decisions using three feature learning methods (principal component analysis (PCA), low rank and sparse matrix decomposition (LSD), and exponential sparse restricted Boltzmann machine (RBM)) and show that the use of features offers improvement in prediction accuracy using over 1 million real passenger vehicle purchase data. We then show that the interpretation and visualization of these feature representations may be used to help augment data-driven design decisions.

UR - http://www.scopus.com/inward/record.url?scp=84971457387&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84971457387&partnerID=8YFLogxK

U2 - 10.1115/1.4033427

DO - 10.1115/1.4033427

M3 - Article

VL - 138

JO - Journal of Mechanical Design - Transactions of the ASME

JF - Journal of Mechanical Design - Transactions of the ASME

SN - 1050-0472

IS - 7

M1 - 071404

ER -