IMPROVING EFFICIENCY OF SUPPORT VECTOR MCLASSIFIER WITH FEATURE SELECTION

บทความในวารสาร

ผู้เขียน/บรรณาธิการ

กลุ่มสาขาการวิจัยเชิงกลยุทธ์

รายละเอียดสำหรับงานพิมพ์

รายชื่อผู้แต่ง: Yaicharoen, Auapong; Hashikura, Kotaro; Kamal, Md Abdus Samad; Yamada, Kou;

ปีที่เผยแพร่ (ค.ศ.): 2022

Volume number: 13

Issue number: 5

หน้าแรก: 479

หน้าสุดท้าย: 486

จำนวนหน้า: 8

นอก: 21852766

URL: https://www.scopus.com/inward/record.uri?eid=2-s2.0-85127328366&doi=10.24507%2ficicelb.13.05.479&partnerID=40&md5=79dadb6f8829099a6ca2447f2a90ffc2

ภาษา: English-Great Britain (EN-GB)

ดูบนเว็บไซต์ของสำนักพิมพ์

บทคัดย่อ

In this paper, a relationship among feature selection, time to train a classi-fier and quality of that classifier is investigated. By choosing different sets of features in a data set to build classifiers, every information from processing time to the quality of each classifier is inspected. The focus is on time used to create a classifier and accuracy of that classifier. Four data sets are used in this investigation. The results from our investigation show that, most of the time, not all features in the data set are necessary to build a good classifier. Those features with higher importance are the ones needed. Also, when an optimum value of threshold is set, a train data set with features that have total value of their importance equal to or better than that threshold can be used to create an equally good quality classifier as the original but required less execution time. © 2022 ICIC International.

คำสำคัญ

Principal component analysis, Support vector classification