Abstract
We propose a novel support vector regression approach called ε-Distance Weighted Support Vector Regression (ε-DWSVR). ε-DWSVR specifically addresses a challenging issue in support vector regression: how to deal with the situation when the distribution of the internal data in the ε-tube is different from that of the boundary data containing support vectors. The proposed ε-DWSVR optimizes the minimum margin and the mean of functional margin simultaneously to tackle this issue. To solve the new optimization problem arising from ε-DWSVR, we adoptdual coordinate descent (DCD) with kernel functions for medium-scale problems and also employ averaged stochastic gradient descent (ASGD) to make ε-DWSVR scalable to larger problems. We report promising results obtained by ε-DWSVR in comparison with five popular regression methods on sixteen UCI benchmark datasets.
Original language | English |
---|---|
Title of host publication | Advances in Knowledge Discovery and Data Mining |
Subtitle of host publication | 22nd Pacific-Asia Conference, PAKDD 2018, Melbourne, VIC, Australia, June 3-6, 2018, Proceedings, Part I |
Editors | Dinh Phung, Vincent S. Tseng, Prof. Geoffrey I. Webb, Bao Ho, Mohadeseh Ganji, Lida Rashidi |
Publisher | Springer International Publishing |
ISBN (Electronic) | 9783319930343 |
ISBN (Print) | 9783319930336 |
DOIs | |
Publication status | Published - 22 Jul 2018 |
Event | PAKDD 2018: The 22nd Pacific-Asia Conference on Knowledge Discovery and Data Mining - Melbourne, Australia Duration: 3 Jun 2018 → 6 Jun 2018 http://prada-research.net/pakdd18/ |
Publication series
Name | Lecture Notes in Artificial Intelligence |
---|---|
Publisher | Springer |
ISSN (Print) | 0302-9743 |
Conference
Conference | PAKDD 2018 |
---|---|
Abbreviated title | PAKDD 2018 |
Country/Territory | Australia |
City | Melbourne |
Period | 3/06/18 → 6/06/18 |
Internet address |
Bibliographical note
We gratefully thank Dr Teng Zhang and Prof Zhi-Hua Zhou for providing thesource code of “LDM”, and their kind technical assistance. We also thank Prof
Chih-Jen Lins team for providing the LIBSVM and LIBLINEAR packages and
their support. This work is supported by the National Natural Science Foundation
of China (Grant Nos.61472159, 61572227) and Development Project of Jilin
Province of China (Grant Nos. 20140101180JC, 20160204022GX, 20180414012G
H). This work is also partially supported by the 2015 Scottish Crucible Award
funded by the Royal Society of Edinburgh and the 2016 PECE bursary provided
by the Scottish Informatics & Computer Science Alliance (SICSA).
Keywords
- regression analysis
- Support Vector Regression
- Distance Weighted Support Vector Regression
- Dual Coordinate Descent
- Averaged Stochastic Gradient Descent