A Financial Distress Prediction Model Based on Sparse Algorithm and Support Vector Machine

Sen Zeng,Yanru Li,Yaqin Li,Wanjun Yang,Chen Chen

doi:10.1155/2020/5625271

Sen Zeng, Yanru Li + Show 3 more

Open Access

https://doi.org/10.1155/2020/5625271

Copy DOI

Abstract

Classification learning is a very important issue in machine learning, which has been widely used in the field of financial distress warning. Some researches show that the prediction model framework based on sparse algorithm has better performance than the traditional model. In this paper, we explore the financial distress prediction based on grouping sparsity. Feature selection of sparse algorithm plays an important role in classification learning, because many redundant and irrelevant features will degrade performance. A good feature selection algorithm would reduce computational complexity and improve classification accuracy. In this study, we propose an algorithm for feature selection classification prediction based on feature attributes and data source grouping. The existing financial distress prediction model usually only uses the data from financial statement and ignores the timeliness of company sample in practice. Therefore, we propose a corporate financial distress prediction model that is better in line with the practice and combines the grouping sparse principal component analysis of financial data, corporate governance characteristics, and market transaction data with support vector machine. Experimental results show that this method can improve the prediction efficiency of financial distress with fewer characteristic variables.

Highlights

In recent years, machine learning algorithms have been widely used in the field of corporate financial distress prediction
An support vector machine (SVM) model based on sparse principal component analysis (GSPCA-SVM) is proposed to deal with financial distress prediction
In the feature selection stage of the original dataset, we propose a method to group the features according to data sources and financial statement analysis. e purpose of this method is to investigate whether the predictive performance of the model can be improved by selecting fewer, relatively more important variables from each information feature category

Summary

Introduction

Machine learning algorithms have been widely used in the field of corporate financial distress prediction. In the field of financial distress prediction, multiple feature selection methods are proposed, such as rough set method, LASSO method, wrapper, and filter [3,4,5] Most of these approaches fail to take into account the attributes and data sources of individual features and the different effects they may have on the tag. Sparse principal component analysis is used to screen the characteristic indexes of each group; a new dataset is formed and substituted into support vector machine (SVM) for classification and prediction.

Related Work

Grouping Sparse PCA-SVM Method

Application

Conclusion

Disclosure