Abstract—Predicting the financial failure performs an even more significant character in the sustainable existence of enterprises for developing countries. A new risk rating technique based on the probability of default (PD) and order statistics (OS) is established to classify listed companies into two categories according to their financial risks. In the present study, the linear kernel function was united with biorthogonal wavelet kernel function to construct a novel biorthogonal hybrid kernel function. Additionally, the probability of default (PD) and Gray relational analysis (GRA) based new feature weighted approach is established. Grey relational degrees (GRD) between PD and financial indicators are the feature weights (FWOCSVM) on account of that PD can provide effective predicting information for the financial crisis of the listed companies. The financial distress was predicted among financially stable and distressed companies by using feature weighted one-class support vector machine based on the probability of default. The results from collected data of listed companies in Karachi Stock Exchange (KSE), Karachi, Pakistan demonstrated adequate performance by using newly developed approach.

Index Terms—Probability of default (PD), support vector machine (SVM), order statistics (OS), FWOCSVM, grey relational degrees (GRD), financial failure.

Bilal Ahmed Khan, Longsheng Cheng are with School of Economics and Management, Nanjing University of Science and Technology, Nanjing, China (email: cheng_longsheng@163.com).
Haris Ahmed is with Institute of Business Administration, University of Sindh, Jamshoro 76090, Sindh, Pakistan.
Muddassar Sarfraz is with Department of Management and HR, Business School, Hohai University, Nanjing China.

Cite: Bilal Ahmed Khan, Longsheng Cheng, Haris Ahmed, Muddassar Sarfraz, "Predicting Financial Failure by Support Vector Machine and Probability of Default of Enterprises in a Developing Country," International Journal of e-Education, e-Business, e-Management and e-Learning vol. 9, no. 4, pp. 285-295, 2019.