Financial distress prediction using the hybrid associative memory with translation

doi:10.1016/j.asoc.2016.04.005

Applied Soft Computing

Volume 44, July 2016, Pages 144-152

https://doi.org/10.1016/j.asoc.2016.04.005 Get rights and content

Highlights

•
We explore the hybrid associative memory with translation for default prediction.
•
We analyze the behavior of this neural network under the presence of class imbalance.
•
We study how the class overlapping affects the performance of the associative memory.
•
We compare its performance with that of other prediction models.
•
The associative memory is the best model, especially to predict the default cases.

Abstract

This paper presents an alternative technique for financial distress prediction systems. The method is based on a type of neural network, which is called hybrid associative memory with translation. While many different neural network architectures have successfully been used to predict credit risk and corporate failure, the power of associative memories for financial decision-making has not been explored in any depth as yet. The performance of the hybrid associative memory with translation is compared to four traditional neural networks, a support vector machine and a logistic regression model in terms of their prediction capabilities. The experimental results over nine real-life data sets show that the associative memory here proposed constitutes an appropriate solution for bankruptcy and credit risk prediction, performing significantly better than the rest of models under class imbalance and data overlapping conditions in terms of the true positive rate and the geometric mean of true positive and true negative rates.

Graphical abstract

Introduction

A large number of techniques have been developed to help decision-makers and analysts in predicting financial distress. Traditionally, decisions on credit risk of a corporate borrower were exclusively based upon subjective judgments made by human experts, using past experiences and some guiding principles [69]. However, two major problems with this approach are the difficulty to make consistent estimates and the fact that it tends to be reactive rather than predictive. The world financial crisis has led to increasing attention of banks and financial institutions on this question because of its significant impact on the decisions made [14], resulting in the development of numerous techniques to face the important challenge of credit risk and bankruptcy prediction from financial ratios using mathematical models. From the pioneer work by Altman [7], based on multivariate discriminant analysis, a variety of statistical and operations research methods have subsequently been applied to credit risk and bankruptcy prediction, including linear and logistic regression, multivariate adaptive regression splines, survival analysis, linear and quadratic programming, and multiple criteria programming. Most of these techniques typically rely on the assumptions of linear separability, multivariate normality and independence of the predictive variables, but they are very often violated in real-life problems [25], [34], [55].

Popular computational intelligence tools such as decision trees, neural networks, support vector machines, fuzzy systems, rough sets, artificial immune systems, and evolutionary algorithms are techniques that can deal with non-linearity. Besides, these methods are highly capable of extracting meaningful information from imprecise data and detecting trends that are too complex to be discovered by either humans or conventional systems. Despite various studies have concluded that no technique is clearly superior to other competing algorithms because it depends on the characteristics of the problem analyzed [13], [15], [16], different neural network architectures have shown good performance in comparison to other methods for a range of financial applications [10], [19], [48], [53], [78]. However, when the number of examples is relatively small, several works have demonstrated that the accuracy and generalization performance of a support vector machine (SVM) is usually better than that of statistical and other soft computing techniques [23], [24], [65], [67]. While typical neural networks used in this context are the multi-layer perceptron (MLP), the radial basis function (RBF) and the probabilistic or Bayesian network (BN), other neural models such as the associative memories have not been explored as yet.

The ability of human brain to make associations from partial information has historically attracted great interest among researchers, leading to a variety of theoretical neural networks that act as associative memories. An associative memory [39] is an early type of artificial neural network that relates an input vector x with an output vector y. The functionality of associative memories is reached in two phases: learning and recall. The learning process consists of building a connection matrix W with a value for each association (x^k, y^k). In the recall phase, an output vector y, which corresponds to the most similar to the input vector x, is obtained from the associative memory. These models are powerful computational tools due to their conceptual and implementational simplicity, their strong mathematical foundation, and their capability of storing huge amounts of data that allow to properly recover the most similar patterns to an input vector with low computational efforts [77].

Representative examples of associative memories are lernmatrix [66], the linear associator [8], [38], the Moore–Penrose generalized inverse associative memory [40], the Hopfield network [28], the bidirectional associative memory [41], the fuzzy associative memory [42], the morphological associative memory [58], and the alpha–beta associative memory [2]. Some of these models have been used to solve very different problems. Sabourin and Mitiche [59] developed a Kohonen associative memory with selective multiresolution for OCR. A fuzzy associative memory was introduced to determine rock types from well-log signatures [17]. The bidirectional associative memory networks were used to find the relations between various cancers and elemental contents in serum samples with the aim of diagnosing cancer [81]. A hybrid classifier based on self-organizing maps and associative memories was designed for speaker recognition [31]. Zhang et al. [79] proposed a modular face recognition scheme by combining the wavelet subband representations and kernel associative memories. An associative memory based on the restricted Coulomb energy was also applied to human face recognition [49]. Namba and Zhang [50] devised an associative memory to recognize Braille images. A novel system for medical diagnosis based on associative memories was proposed by Aldape-Pérez et al. [5]. Itkar and Kulkarni [32] developed an efficient algorithm for mining frequent patterns using an auto-associative memory.

Apart from the associative memories just mentioned, Santiago-Montero [63] introduced the hybrid associative classifier and its extension, the hybrid associative classifier with translation (HACT). Both these associative memories are based on the learning phase of the linear associator and the recall phase of the Steinbuch's lernmatrix. This paper applies the HACT neural network to decision making problems for financial distress prediction and presents an empirical comparison with other popular prediction methods. To the best of our knowledge, this model has not been used for classification purposes, and even less in the context of finance and management. The aim of this paper therefore is four-fold:

1.
To explore the capability of the HACT model in the prediction of bankruptcy and credit risk.
2.
To analyze the behavior of this neural network under the presence of imbalance in class distribution, which constitutes a data complexity often neglected in financial applications.
3.
To investigate how the class overlapping affects the performance of the associative memory.
4.
To compare the performance of HACT with that of other prediction techniques.

From now on, the paper is organized as follows. Section 2 provides a review of works related to neural networks used for corporate bankruptcy and credit risk prediction. Section 3 introduces the fundamental concepts of the associative memories and describes the bases of the HACT model. The experimental set-up and databases are given in Section 4, while the results are discussed in Section 5. Finally, Section 6 presents the concluding remarks and outlines some directions for future research.

Section snippets

A review of neural networks applied to financial distress prediction

From the beginning of the 1990s, the development of artificial neural network technologies for bankruptcy and credit risk prediction problems has been the subject of considerable attention and research efforts. The first reference to using neural networks can be found in the paper by Odom and Sharda [51], showing that a three-layer feed-forward perceptron is more accurate and robust than multi-variate discriminant analysis. After this seminal work, many other studies have proposed the use of

Hybrid associative classifier with translation

In its most general form, an associative memory is a content-addressable neural network based on matrix algebra [39], [57] that maps input patterns (examples) to output patterns by using the p different associated pattern pairs (x^k, y^k) stored during the learning phase. The associative memory takes the form of a connection weight matrix $W = {[w_{i, j}]}_{m \times n}$ generated from a finite set of p encoded associations, called fundamental set of associations, {(x^μ, y^μ)|μ = 1, 2, …, p}, where $x^{μ} \in ℝ^{n}$ are the

Experimental set-up

Nine data sets related to bankruptcy/creditworthiness have been employed in order to make a comprehensive comparison of the HACT model with four well-known neural networks (MLP, RBF, BN and the voted perceptron, VP), whose architectures and parameter settings are reported in Table 1. In addition, an SVM with a linear kernel (widely acknowledged as one of the best soft computing techniques) and the logit model (a classical econometric method) have also been included in this study. Note that,

Experimental results and discussion

Table 4 reports the true positive rate averaged across the 10 runs for each database, the average values across all the databases and the Friedman's average rank for each neural network approach (the one with the lowest average rank has to be deemed as the best solution). The values for the best performing method in each database are underlined. Based on the Friedman's average ranks, the results reveal that the HACT model corresponds to the algorithm with the best performance, followed by MLP

Conclusions and future work

From the first works in the beginning of the 1990s, the artificial neural networks emerged as an effective method for bankruptcy and credit risk prediction. They differ from classical financial prediction systems, such as the models based on statistical techniques, mainly in their black-box nature and in the assumption of a non-linear relation among variables. In this paper, the hybrid associative memory with translation has been explored and compared to other well-known neural models (MLP,

Acknowledgments

This work has partially been supported by the Mexican CONACYT through the Postdoctoral Fellowship Program [232167], the Spanish Ministry of Economy [TIN2013-46522-P], the Generalitat Valenciana [PROMETEOII/2014/062] and the Mexican PRODEP [DSA/103.5/15/7004]. We would like to thank the Reviewers for their valuable comments and suggestions, which have helped to improve the quality of this paper substantially.

References (81)

H. Abdou et al.
Neural nets versus conventional techniques in credit scoring in Egyptian banking
Expert Syst. Appl.
(2008)
M. Aldape-Pérez et al.
An associative memory approach to medical decision support systems
Comput. Methods Prog. Biomed.
(2012)
E. Alfaro et al.
Bankruptcy forecasting: an empirical comparison of AdaBoost and neural networks
Decis. Support Syst.
(2008)
J.A. Anderson
A simple neural network generating an interactive memory
Math. Biosci.
(1972)
E. Angelini et al.
A neural network approach for credit risk evaluation
Q. Rev. Econ. Financ.
(2008)
I. Brown et al.
An experimental comparison of classification algorithms for imbalanced credit scoring data sets
Expert Syst. Appl.
(2012)
W.S. Chen et al.
Using neural networks and data mining techniques for the financial distress prediction model
Expert Syst. Appl.
(2009)
C.B. Cheng et al.
Financial distress prediction by a radial basis function network with logit analysis learning
Comput. Math. Appl.
(2006)
C.L. Chuang et al.
Constructing a reassigning credit scoring model
Expert Syst. Appl.
(2009)
V.S. Desai et al.
A comparison of neural networks and linear scoring models in the credit union environment
Eur. J. Oper. Res.
(1996)

Y. Ding et al.

Forecasting financial condition of Chinese listed companies based on support vector machine

Expert Syst. Appl.

(2008)

N.C. Hsieh

Hybrid mining approach in the design of credit scoring models

Expert Syst. Appl.

(2005)

C. Hung et al.

A selective ensemble based on expected probabilities for bankruptcy prediction

Expert Syst. Appl.

(2009)

A. Khashman

Neural networks for credit risk evaluation: investigation of different neural models and learning schemes

Expert Syst. Appl.

(2010)

A. Khashman

Credit risk evaluation using neural networks: emotional versus conventional models

Appl. Soft Comput.

(2011)

R.C. Lacher et al.

A neural network for classifying the financial health of a firm

Eur. J. Oper. Res.

(1995)

T.S. Lee et al.

A two-stage hybrid credit scoring model using artificial neural networks and multivariate adaptive regression splines

Expert Syst. Appl.

(2005)

T.S. Lee et al.

Credit scoring using the hybrid neural discriminant technique

Expert Syst. Appl.

(2002)

S. Lessmann et al.

Benchmarking state-of-the-art classification algorithms for credit scoring: an update of research

Eur. J. Oper. Res.

(2015)

T.H. Lin

A cross model study of corporate financial distress prediction in Taiwan: multiple discriminant analysis, logit, probit and neural networks models

Neurocomputing

(2009)

P.C. Pendharkar

A threshold-varying artificial neural network approach for classification and its application to bankruptcy prediction problem

Comput. Oper. Res.

(2005)

V. Ravi et al.

Soft computing system for bank performance prediction

Appl. Soft Comput.

(2008)

V. Ravi et al.

Threshold accepting trained principal component neural network and feature subset selection: application to bankruptcy prediction in banks

Appl. Soft Comput.

(2008)

M. Sabourin et al.

Modeling and classification of shape using a Kohonen associative memory with selective multiresolution

Neural Netw.

(1993)

C. Serrano-Cinca et al.

Partial least square discriminant analysis for bankruptcy prediction

Decis. Support Syst.

(2013)

K.S. Shin et al.

An application of support vector machines in bankruptcy prediction model

Expert Syst. Appl.

(2005)

J. Sun et al.

Financial distress prediction using support vector machines: ensemble vs. individual

Appl. Soft Comput.

(2012)

C.F. Tsai et al.

A comparative study of classifier ensembles for bankruptcy prediction

Appl. Soft Comput.

(2014)

C.F. Tsai et al.

Using neural network ensembles for bankruptcy prediction and credit scoring

Expert Syst. Appl.

(2008)

D. West

Neural network credit scoring models

Comput. Oper. Res.

(2000)

D. West et al.

Neural network ensemble strategies for financial decision applications

Comput. Oper. Res.

(2005)

L. Yu et al.

Credit risk assessment with a multistage neural network ensemble learning approach

Expert Syst. Appl.

(2008)

G. Zhang et al.

Artificial neural networks in bankruptcy prediction: general framework and cross-validation analysis

Eur. J. Oper. Res.

(1999)

Z. Zhang et al.

Classification of cancer patients based on elemental contents of serums using bidirectional associative memory networks

Anal. Chim. Acta

(2001)

M.E. Acevedo-Mosqueda et al.

Alpha–beta bidirectional associative memories: theory and applications

Neural Process. Lett.

(2007)

J. Alcalá-Fdez et al.

KEEL data-mining software tool: data set repository, integration of algorithms and experimental analysis framework

J. Mult. Valued Log. Soft Comput.

(2011)

J. Alcalá-Fdez et al.

KEEL: a software tool to assess evolutionary algorithms to data mining problems

Soft Comput.

(2009)

E. Altman

Financial ratios, discriminant analysis, and the prediction of corporate bankruptcy

J. Financ.

(1968)

A.F. Atiya

Bankruptcy prediction for credit risk using neural networks: a survey and new results

IEEE Trans. Neural Netw.

(2001)

J. Baek et al.

Bankruptcy prediction for credit risk using an auto-associative neural network in Korean firms

Cited by (70)

Bankruptcy prediction with low-quality financial information
2024, Expert Systems with Applications
The corporate bankruptcy prediction literature has traditionally relied on data from public, audited companies. However, the vast majority of firms worldwide are privately-held and lack the same level of scrutiny over their financial statements. As a result, these businesses usually produce less accurate and transparent accounting reports. Our research problem is to address this gap: how stakeholders deal with these less reliable information? Using a novel dataset of 503 private firms that filed for reorganization in Brazil between 2007 and 2020, we found that financial ratios had a significantly lesser effect on explaining default and bankruptcy than what previous research suggested, due in part to the lower information content in the accounting statements within our database. Instead, lenders seem to focus on harder-to-conceal variables, such as collateralizable assets, as well as on institutional factors, like proxies of financial statement quality. There is also concerning evidence that specialized attorneys can ”work the system” in favor of distressed companies regardless of their financial fundamentals. Additionally, we found that machine learning models outperformed traditional statistical ones in different sorts of metrics, corroborating the literature on the superior performance of non-linear approaches on datasets having synergistic causality among its features.
A novel federated learning approach with knowledge transfer for credit scoring
2024, Decision Support Systems
The expanding availability of data in the financial sector promises to take the performance of machine learning models to a new level. However, given the high business value and confidentiality of credit data, the integration of datasets from multiple institutions for credit scoring modeling may result in privacy leakage. Consequently, in this paper, a horizontal federated learning paradigm is used to protect the local private data of each participant and collaborate to train a powerful shared global model. However, in the collaborative training process, heterogeneous data distributions can result in insufficient learning of the model. To overcome this issue, we propose the federated knowledge transfer (FedKT) method, which exploits the advantages of fine-tuning and knowledge distillation to effectively extract generic and specific knowledge from the early layers and outputs of the global model, respectively, thus improving the learning performance of the local models. We adopt five credit datasets and four performance measures to demonstrate the effectiveness of our proposed method. The experimental results show that the proposed method can securely utilize credit data from different parties to improve the performance of the credit scoring model. This also supports the potential of our proposed method for further applications in credit scoring.
Long-horizon predictions of credit default with inconsistent customers
2024, Technological Forecasting and Social Change
We developed a decision support framework for default predictions that addresses two common issues: inconsistent customers and predictions of future defaults. We developed a $T - m$ default prediction model using multivariate adaptive regression splines to address the methodological challenges. We confirm that this model outperforms typical approaches in terms of default prediction accuracy. Furthermore, an empirical application of our new framework involving unique data on defaults among Chinese-listed companies yields several substantive insights. Owing to the high interpretability of our predictions, we identify certain industry sectors that should receive high (and low) credit risk assessments. In addition, our research has important implications for the investment decisions of financial institutions and investors and government regulations.
Predicting and interpreting financial distress using a weighted boosted tree-based tree
2022, Engineering Applications of Artificial Intelligence
Financial distress prediction aims at providing an early warning solution of financial distress to help business participants, investors, and regulators to achieve better profit growth and financial risk management. Extreme gradient boosting (XGBoost), has been recognized as a favorable competitor compared with machine learning-based individual classifiers. However, its commercial value for FDP is hindered by two reasons. First, FDP is a classical imbalance issue, traditional XGBoost is considered a cost-insensitive approach that yields skew-sensitive FDP results. Second, XGBoost is a complex ensemble approach that faces the performance-interpretability dilemma, making the decision logic of XGBoost cannot be easily understood. To solve the above limitations, in this study, we first focus on addressing the imbalance issue in FDP by introducing a weighted cost-sensitive XGBoost, reducing the error of misclassifying financial distress firms. Next, we merge the decision rules extracted from the optimized weighted XGBoost to reconstruct a new tree as the approximation of the cost-sensitive ensemble model, making the proposed weighted XGBoost-based tree (XGBoost-W-BT) an accurate and interpretable solution for imbalanced FDP. Experimental results on a Chinese FDP dataset collected from China Security Market Accounting Research Database (CSMARD) showed that XGBoost-W-BT can be an alternative to weighted XGBoost to predict financial distress at an early stage. Besides, the transparent tree-based structure provides an explicit explanation to help industry participants and regulators make scientific policies, guiding investors to make rational investments.
Assessing credit risk of commercial customers using hybrid machine learning algorithms
2022, Expert Systems with Applications
Given the large amount of customer data available to financial companies, the use of traditional statistical approaches (e.g., regressions) to predict customers’ credit scores may not provide the best predictive performance. Machine learning (ML) algorithms have been explored in the credit scoring literature to increase predictive power. In this paper, we predict commercial customers’ credit scores using hybrid ML algorithms that combine unsupervised and supervised ML methods. We implement different approaches and compare the performance of the hybrid models to that of individual supervised ML models. We find that hybrid models outperform their individual counterparts in predicting commercial customers’ credit scores. Further, while the existing literature ignores past credit scores, we find that the hybrid models’ predictive performance is higher when these features are included.
A novel ensemble feature selection method by integrating multiple ranking information combined with an SVM ensemble model for enterprise credit risk prediction in the supply chain
2022, Expert Systems with Applications
Citation Excerpt :
The cost of incorrectly classifying minority class instances into the majority class is very high (Sun et al., 2020). At present, there are three main kinds of solutions to solve the class imbalance problem: the data-level method (Cordón et al., 2018; Gao et al., 2014; Li et al., 2018; Zhang & Li, 2014), the algorithm-level method (Cleofas-Sánchez et al., 2016; Raghuwanshi & Shukla, 2018) and the ensemble learning method (Shen et al., 2020; Wang & Sun, 2021). Scholars are devoting increasing attention to developing new models by combining the data-level method and the ensemble learning method.
Enterprise credit risk prediction in the supply chain context is an important step for decision making and early credit crisis warnings. Improving the prediction performance of this task is an academic and industrial focus. Feature selection and class imbalance can affect prediction performance: redundant and irrelevant features increase the learning difficulty of the prediction model, cause overfitting and reduce prediction performance, whereas class imbalance, with many fewer minority class instances than majority class instances, may cause model failure. Herein, a sequence backward feature selection algorithm based on ranking information (SBFS-RI) and a novel ensemble feature selection method integrating multiple ranking information (FS-MRI) are proposed. The FS-MRI method can realize the automatic threshold function while considering the model performance and then output the best and a more stable feature subset. In addition, an SVM ensemble model with an artificial imbalance rate (SVME-AIR) is proposed to solve the class imbalance problem and realize the effective combination of under-sampling technology and the AdaBoost ensemble method for the first time. Finally, FS-MRI and SVME-AIR are combined through a two-stage model design. The hybrid model can effectively solve the feature selection and class imbalance problems for enterprise credit risk prediction in the supply chain context. Supply chain data of Chinese listed enterprises shows that the FS-MRI method outperforms nine other feature selection methods and provides more robust and efficient feature subsets. The SVME-AIR model has higher AUC and KS values than other ensemble models and single classifiers. When combined, the two methods achieve the best prediction performance, with maximum AUC and KS values of 0.8772 and 0.6363, respectively.

View all citing articles on Scopus

View full text

Financial distress prediction using the hybrid associative memory with translation

Highlights

Abstract

Graphical abstract

Introduction

Section snippets

A review of neural networks applied to financial distress prediction

Hybrid associative classifier with translation

Experimental set-up

Experimental results and discussion

Conclusions and future work

Acknowledgments

Expert Syst. Appl.

Comput. Methods Prog. Biomed.

Decis. Support Syst.

Math. Biosci.

Q. Rev. Econ. Financ.

Expert Syst. Appl.

Expert Syst. Appl.

Comput. Math. Appl.

Expert Syst. Appl.

Eur. J. Oper. Res.

Expert Syst. Appl.

Expert Syst. Appl.

Expert Syst. Appl.

Expert Syst. Appl.

Appl. Soft Comput.

Eur. J. Oper. Res.

Expert Syst. Appl.

Expert Syst. Appl.

Eur. J. Oper. Res.

Neurocomputing

Comput. Oper. Res.

Appl. Soft Comput.

Appl. Soft Comput.

Neural Netw.

Decis. Support Syst.

Expert Syst. Appl.

Appl. Soft Comput.

Appl. Soft Comput.

Expert Syst. Appl.

Comput. Oper. Res.

Comput. Oper. Res.

Expert Syst. Appl.

Eur. J. Oper. Res.

Anal. Chim. Acta

Alpha–beta bidirectional associative memories: theory and applications

Neural Process. Lett.

KEEL data-mining software tool: data set repository, integration of algorithms and experimental analysis framework

J. Mult. Valued Log. Soft Comput.

KEEL: a software tool to assess evolutionary algorithms to data mining problems

Soft Comput.

Financial ratios, discriminant analysis, and the prediction of corporate bankruptcy

J. Financ.

Bankruptcy prediction for credit risk using neural networks: a survey and new results

IEEE Trans. Neural Netw.

Bankruptcy prediction for credit risk using an auto-associative neural network in Korean firms