DETERMINATION OF AN OPTIMAL PIPELINE FOR IMBALANCED CLASSIFICATION: PREDICTING POTENTIAL CUSTOMER COMPLAINTS TO A TEXTILE MANUFACTURER
DOI:
https://doi.org/10.23055/ijietap.2020.27.5.6757Keywords:
textile manufacturer, customer complaint, class-imbalanced problem, classification, multiple response design of experimentsAbstract
There is an urgent need to reduce customer complaints because they damage reputations and incur losses. This study predicts the likelihood of complaint about a new production order using its intrinsic features. Customer complaints, however, are relatively rare, creating a serious class-imbalanced problem when training a classifier. To overcome this problem, we use a pipeline including the upsampling, the hyper-parameter generation, the classifier, and the evaluation metric. As each strategy involves different tricks in the pipeline, we use the design of experiments (DOE) concept to find, automatically, a suitable combination. A multi-response DOE is used to maximize balanced accuracy and minimize overfitting during training. The experimental results showed that the balanced accuracy of the proposed method for the testing dataset was about 23.6% better than those of the base classifiers and about 7% better than those of the current state-of-the-art methods.
Published
How to Cite
Issue
Section
License
The Author(s) must formally transfer each article's copyright before publication in the INTERNATIONAL JOURNAL OF INDUSTRIAL ENGINEERING. Such transfer enables the Journal to defend itself against plagiarism and other forms of copyright infringement. Your cooperation is appreciated.
You agree that the copyright of your article to be published in the INTERNATIONAL JOURNAL OF INDUSTRIAL ENGINEERING - THEORY, APPLICATIONS, AND PRACTICE is hereby transferred, throughout the World and for the full term and all extensions and renewals thereof, to INTERNATIONAL JOURNAL OF INDUSTRIAL ENGINEERING - THEORY, APPLICATIONS, AND PRACTICE.
The Author(s) reserve(s): (a) the trademark rights and patent rights, if any, and (b) the right to use all or part of the information contained in this article in future, non-commercial works of the Author's own, or, if the article is a "work-for-hire" and made within the scope of the Author's employment, the employer may use all or part of the information contained in this article for intra-company use, provided the usual acknowledgments are given regarding copyright notice and reference to the original publication.
The Author(s) warrant(s) that the article is Author's original work and has not been published before. If excerpts from copyrighted works are included, the Author will obtain written permission from the copyright owners and credit the article's sources.
The author also warrants that the article contains no libelous or unlawful statements and does not infringe on the rights of others. If the article was prepared jointly with other Author(s), the Author agrees to inform the co-Author(s) of the terms of the copyright transfer and to sign on their behalf; or in the case of a "work-for-hire," the employer or an authorized representative of the employer.
The journal does not provide the author copy of the final paper when it is published. The author(s) can make(s) a subscription to INTERNATIONAL JOURNAL OF INDUSTRIAL ENGINEERING - THEORY, APPLICATIONS, AND PRACTICE if they want to get the final paper that has already been published.
The journal is registered with the Library of Congress (ISSN # 1943-670X). All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted in any form or by any means, electronic, mechanical, photocopying, recording, or otherwise, without the prior written permission of the journal.
The author reserves patent and trademark rights and the right to use all or part of the information contained in the article in future non-commercial works.
