A kernel fuzzy twin SVM model for early warning systems of extreme financial risks

• DOI: http://doi.org/10.1002/ijfe.1858
• Author: Xun Huang,Fanyong Guo
• Date: 01 January 2021
• Published date: 01 January 2021

RESEARCH ARTICLE

A kernel fuzzy twin SVM model for early warning systems

of extreme financial risks

Xun Huang

1,2

| Fanyong Guo

3

1

Business School, Chengdu University,

Chengdu, China

2

Institute of Chinese Financial Studies,

Southwest University of Finance and

Economics, Chengdu, China

3

School of Finance, Shanghai University

of Finance and Economics, Shanghai,

China

Correspondence

Fanyong Guo, School of Finance,

Shanghai University of Finance and

Economics, Shanghai 20043, China.

Email: guofanyong@163.sufe.edu.cn

Funding information

National Social Science Fund of China,

Grant/Award Number: 17BGL231;

Southwestern University of Finance and

Economics, Grant/Award Number:

JBK1805003

Abstract

It is an important component of risk management in financial markets to

develop an early warning systems (EWS) for extreme financial risk. In this

paper, we establish a novel EWS called kernel fuzzy twin support vector

machine (KFT-SVM). Unlike T-SVM, KFT-SVM can deal with the noises and

outliners in dataset and the fuzzy dataset with a lot of potential uncertain but

important factors in financial markets by introducing the fuzzy approach. More

importantly, the introduced kernel method can aid the fuzzy approach to

achieve more valuable fuzzy memberships by transporting dataset from the

input space to the kernel space and further improve the generalization perfor-

mance of T-SVM. Computational comparisons of KFT-SVM against SVM,

T-SVM and FT-SVM indicate the significant superiority of our proposed KFT-

SVM. Furthermore, we have investigated the favourable ability of KFT-SVM for

overcoming the class imbalance problem by comparison with that combined

with the resampling method of the synthetic minority over-sampling technique

(SMOTE). The experimental result showsthat our proposed KFT-SVM can effec-

tively overcome the class imbalance problem.

KEYWORDS

early warning system, extreme financial risks, fuzzy approach, kernel method, twin support

vector machine

1|INTRODUCTION

In recent years, extreme financial risks represented by the

Asian crisis and the subprime crisis have threatened the

stability of financial markets all over the world

(Cumperayot & Kouwenberg, 2013). In order to prepare

against extreme financial risks effectively, it is quite neces-

sary to establish a proper early warning system (EWS) to

predict potential extreme financial risks (Lang &

Schmidt, 2016). In fact, it is essential for EWS to predict

extreme financial risks accurately via classification

approaches with a perfect performance (Oh, Kim, &

Kim, 2006; Oh, Kim, Kim, & Lee, 2006). For this, a large

number of researchers have devoted themselves to devel-

oping various classification approaches, especially artificial

intelligence algorithms, for achieving a desirable accuracy

of EWS. Kim, Hwang, and Lee (2004) and Kim, Oh, Sohn,

and Hwang (2004) suggested that artificial neural net-

works (ANN) may predict financial risks effectively. How-

ever, Oh and Kim (2007) proposed that case-based

reasoning (CBR) has competitive edge over ANN. Further,

Ahn, Oh, Kim, and Kim (2011) developed support vector

machine (SVM) for extending the EWS classification

approach, verifying that SVM is superior to other classifi-

cation approaches including ANN and CBR.

Declaration: The manuscript contains original unpublished work and

has not been submitted for publication elsewhere at the same time.

Received: 14 November 2018 Revised: 26 October 2019 Accepted: 18 June 2020

DOI: 10.1002/ijfe.1858

Int J Fin Econ. 2021;26:1459–1468. wileyonlinelibrary.com/journal/ijfe © 2020 John Wiley & Sons, Ltd. 1459