Can machine learning approaches predict corporate bankruptcy? Evidence from a qualitative experimental design

Bankruptcy prediction has received a growing interest in corporate finance and risk management recently. Although numerous studies in the literature have dealt with various statistical and artificial intelligence classifiers, their performance in credit risk forecasting needs to be further scrutinized compared to other methods. In the spirit of Chen, Härdle and Moro (2011, Quantitative Finance), we design an empirical study to assess the effectiveness of various machine learning topologies trained with big data approaches and qualitative, rather than quantitative, information as input variables. The experimental results from a ten-fold cross-validation methodology demonstrate that a generalized regression neural topology yields an accuracy measurement of 99.96%, a sensitivity measure of 99.91% and specificity of 100%. Indeed, this specific model outperformed multi-layer back-propagation networks, probabilistic neural networks, radial basis functions and regression trees, as well as other advanced classifiers. The utilization of advanced nonlinear classifiers based on big data methodologies and machine learning training generates outperforming results compared to traditional methods for bankruptcy forecasting and risk measurement.     

Research Opportunities for Neural Networks: The Case for Credit

This article identifies research opportunities in the use of artificial neural networks in credit scoring and related business intelligence situations, particularly as they have been emerging in the global economy. In the literature review, particular attention is paid to commercial lending credit risk assessment and consumer credit scoring. Investors and auditors need models that can predict whether a customer will stay viable. Lenders must manage their credit risk to maximize profits and cash flow, while minimizing losses. As the global economic recession continues, investors are tightening their investment belts and need models that help them make better investment decisions, while lenders must strengthen lending practices and better identify both good and bad credit risks. Artificial neural networks may help firms improve their credit model development, and thereby their credit decisions and profitability. Such technology may also help improve development in emerging economies. Copyright © 2016 John Wiley & Sons, Ltd.     

基于RBF网络的商业银行信用风险控制研究

对信用风险的有效控制与管理,在现代商业银行日常运行过程中具有举足轻重的地位。基于信用风险系统是一个高度复杂的非线性动态系统,利用神经网络的自适应学习、并行分布处理和较强的鲁棒性及容错性等特性,建立基于RBF神经网络的信用风险预测控制模型,从理论上探寻信用风险非线性智能控制。仿真试验表明,信用风险度能被控制在以最佳风险度为中心的一定范围内。因此,该预测控制系统适合于商业银行信用风险的控制。     

失信风险传染会影响债券定价吗？——基于担保网络大数据的实证研究

在债券定价研究中不仅应该考虑企业自身的信用风险,还应该考虑相关网络组织的传染风险。本文基于43万笔非金融企业间的担保数据,构建了企业信用担保网络,发现失信风险作为一种广义的信用风险,在担保网络中具有传染效应,该传染效应能够影响债券的信用利差。企业的失信行为产生了三类传染效应,一是直接传染效应,无论是发债主体的担保人出现失信行为,还是被担保人出现失信行为,都会引起发债主体的信用利差上升;二是局部感染效应,如果局部担保网络中失信主体的占比提升,可能引起投资者对发债主体的“团体处罚”,导致信用利差上升;三是全局扩散效应,失信信息沿担保网络向整个市场扩散,导致债券信用利差上升。从企业所有制来看,民营企业主要受微观的直接传染效应和中观的局部感染效应影响;而国有企业主要受全局扩散效应影响;被担保人的失信风险对两类企业都有显著影响。失信风险传染效应会降低企业的再融资能力,其中局部感染效应导致企业次年的借款融资额下降,全局扩散效应导致企业次年的债券融资额下降。     

Predicting credit card fraud with Sarbanes‐Oxley assessments and Fama‐French risk factors

This research developed and tested machine learning models to predict significant credit card fraud in corporate systems using Sarbanes‐Oxley (SOX) reports, news reports of breaches and Fama‐French risk factors (FF). Exploratory analysis found that SOX information predicted several types of security breaches, with the strongest performance in predicting credit card fraud. A systematic tuning of hyperparamters for a suite of machine learning models, starting with a random forest, an extremely‐randomized forest, a random grid of gradient boosting machines (GBMs), a random grid of deep neural nets, a fixed grid of general linear models where assembled into two trained stacked ensemble models optimized for F1 performance; an ensemble that contained all the models, and an ensemble containing just the best performing model from each algorithm class. Tuned GBMs performed best under all conditions. Without FF, models yielded an AUC of 99.3% and closeness of the training and validation matrices confirm that the model is robust. The most important predictors were firm specific, as would be expected, since control weaknesses vary at the firm level. Audit firm fees were the most important non‐firm‐specific predictors. Adding FF to the model rendered perfect prediction (100%) in the trained confusion matrix and AUC of 99.8%. The most important predictors of credit card fraud were the FF coefficient for the High book‐to‐market ratio Minus Low factor. The second most influential variable was the year of reporting, and third most important was the Fama‐French 3‐factor model R² – together these described most of the variance in credit card fraud occurrence. In all cases the four major SOX specific opinions rendered by auditors and the signed SOX report had little predictive influence.     

An Empirical Analysis of Personal Bankruptcy and Delinquency

This article uses a new dataset of credit card accounts to analyzecredit card delinquency, personal bankruptcy, and the stabilityof credit risk models. We estimate duration models for defaultand assess the relative importance of different variables inpredicting default. We investigate how the propensity to defaulthas changed over time, disentangling the two leading explanationsfor the recent increase in default rates—a deteriorationin the risk composition of borrowers versus an increase in borrowers'willingness to default due to declines in default costs. Evenafter controlling for risk composition and economic fundamentals,the propensity to default significantly increased between 1995and 1997. Standard default models missed an important time-varyingdefault factor, consistent with a decline in default costs.     

Foreclosure Delay and Consumer Credit Performance

The deep housing market recession from 2008 through 2010 was characterized by a steep rise in number of foreclosures. The average length of time from onset of delinquency through the end of the foreclosure process also expanded dramatically. Although most individuals undergoing foreclosure were experiencing serious financial stress, the extended foreclosure timelines enabled them to live in their homes without making mortgage payments until the end of the foreclosure process, thus providing temporary income and liquidity benefits from lower housing costs. This paper investigates the impact of extended foreclosure timelines on borrower performance with credit card debt. Our results indicate that a longer period of nonpayment of mortgage expenses results in higher cure rates on delinquent credit cards and reduced credit card balances. Thus, foreclosure process delays may have mitigated the impact of the economic downturn on credit card default—suggesting that improvement in credit card performance during the post-crisis period would likely be slowed by the removal of the temporary liquidity benefits as foreclosures reach completion.     

Interpretable credit model development via artificial neural networks

Poor credit granting decisions are coming back to haunt providers of loan finance. Past poor credit granting decisions are in part due to the Equal Credit Opportunity Act (1975). This act requires lenders to explain the decision to grant or refuse credit. As a result, models such as artificial neural networks, which offer improved ability to identify poor credit risks but which do not offer easy explanations of why a loan applicant has scored badly, remain unused. This paper investigates whether these models can be interpreted so that explanations for credit application rejection can be provided. The results indicate that while the artificial neural networks can be used (with caution) to develop credit scoring models, the limitations imposed by the credit granting process make their use unlikely until interpretation techniques are developed that are more robust and that can interpret multiple hidden-layer artificial neural networks. Copyright © 2008 John Wiley & Sons, Ltd.     

基于人工智能的区域现金需求预测模型研究——以南阳市为例

当前,我国非现金支付对现金支付的替代作用日益显著,但现金支付在居民日常支付中仍占据重要地位。在此背景下,如何精确地预测区域现金需求,保证满足区域现金流通需要已成为一项重要课题。通过借鉴国内外机器学习、神经网络等人工智能技术,基于地区现金投放历史数据,使用岭回归、MLP和LSTM三种模型,构建RR-MLP-LSTM加权组合模型实现了区域现金需求预测,对于统筹货币发行工作、积极防范和化解货币供应风险具有较为重要的借鉴意义。     

Additional Credit for Liquidity-Constrained Individuals: High-Interest Consumer Credit in Korea

We find that the delinquency probability on formal sector debts of private loan borrowers in Korea increases from 2.4% to 20% in the first year after the borrowing and to 32% in the second year. This increase happens despite private loan borrowers trying to rebuild their financial health by reducing formal sector debts, credit card cash service balances, and credit card purchases during the post-borrowing period. This limits the possibility of moral hazard driving the results. Private loan amounts are positively associated with the delinquency probability after controlling other commonly used variables, suggesting that they contain additional information on the worsening financial situation of an individual.     

The Importance of Adverse Selection in the Credit Card Market: Evidence from Randomized Trials of Credit Card Solicitations

Analyzing unique data from multiple large‐scale randomized marketing trials of preapproved credit card solicitations by a large financial institution, we find that consumers responding to the lender's inferior solicitation offers have poorer credit quality attributes. This finding supports the argument that riskier type borrowers are liquidity or credit constrained and, thus, have higher reservation loan interest rates. We also find a more severe deterioration ex post in the credit quality of the booked accounts of inferior offer types relative to superior offers. After controlling for a cardholder's observable risk attributes, demographic characteristics, and adverse economic shocks, we find that cardholders who responded to the inferior credit card offers are significantly more likely to default ex post. Our results provide evidence on the importance of adverse selection effects in the credit card market.     

Debit or credit?

Empirical consumer payment price sensitivity has implications for theory, optimal regulation of payment card networks, and business strategy. A critical margin is the price of a credit card charge. A revolver who did not pay her most recent balance in full pays interest; other credit card users do not. I find that revolvers are substantially less likely to incur credit card charges and substantially more likely to use a debit card, conditional on several proxies for transaction demand and tastes. Debit use also increases with credit limit constraints and decreases with credit card possession. Additional results suggest that debit is becoming a stronger substitute for credit over time.     

Interpretable selective learning in credit risk

Forecasting credit default risk has been an important research field for several decades. Traditionally, logistic regression has been widely recognized as a solution because of its accuracy and interpretability. Although complex machine learning models may improve accuracy over simple logistic regressions, their interpretability has prevented their use in credit risk assessment. We introduce a neural network with a selective option to increase interpretability by distinguishing whether linear models can explain the dataset. Our methods are tested on two datasets: 25,000 samples from the Taiwan payment system collected in October 2005 and 250,000 samples from the 2011 Kaggle competition. We find that, for most of samples, logistic regression will be sufficient, with reasonable accuracy; meanwhile, for some specific data portions, a shallow neural network model leads to much better accuracy without significantly sacrificing interpretability.     

信用卡使用意向成因分析及其营销策略研究

本文在计划行为理论的基础上,使用结构方程模型等数理统计方法,探析了消费者信用卡使用意向的影响因素,进而建构了消费者信用卡使用意向模型。研究表明,计划行为理论中信用卡态度和知觉行为控制变量对消费者信用卡使用意向有重要影响。此外,研究还进一步将消费者信用卡态度区分为便利性态度、金钱与信用态度,实证研究显示便利性态度对信用卡使用意向有重要影响,这种区分是合理的和必要的。最后,研究提出实务建议,为中国信用卡业务的完善和发展提供决策依据。     

Feedforward versus recurrent neural networks for forecasting monthly japanese yen exchange rates

Neural networks are a relatively new computer artificial intelligence method which attempt to mimic the brain's problem solving process and can be used for predicting nonlinear economic time series. Neural networks are used to look for patterns in data, learn these patterns, and then classify new patterns and make forecasts. Feedforward neural networks pass the data forward from input to output, while recurrent networks have a feedback loop where data can be fed back into the input at some point before it is fed forward again for further processing and final output. Some have argued that since time series data may have autocorrelation or time dependence, the recurrent neural network models which take advantage of time dependence may be useful. Feedforward and recurrent neural networks are used for comparison in forecasting the Japanese yen/US dollar exchange rate. A traditional ARIMA model is used as a benchmark for comparison with the neural network models.Results for out of sample show that the feedforward model is relatively accurate in forecasting both price levels and price direction, despite being quite simple and easy to use. However, the recurrent network forecast performance was lower than that of the feedforward model. This may be because feed forward models must pass the data from back to forward as well as forward to back, and can sometimes become confused or unstable. Both the feedforward and recurrent models performed better than the ARIMA benchmark model.The author wish to thank the reviewers Drs. Kraft and Radford for their helpful comments.     

基于Logistic回归模型的个人小额贷款信用风险评估及应用

构建二分类 Logistic信用风险评估模型,运用光大银行某分行样本数据,评估商业银行互联网金融个人小额贷款信用风险。结果显示：客户性别、学历、年龄、收入、职业、属地等因素对个人小额贷款信用风险影响显著。其中,年龄、收入、学历等与客户信用等级呈正向关系,女性信用风险显著低于男性,持有信用卡、存贷比越低的客户其信用等级越高;一、二线城市客户的履约率普遍高于县地级市客户的履约率。鉴此,商业银行应对互联网金融个人小额贷款信用风险进行有效规避和分散。     

Making trading decisions for financial‐engineered derivatives: a novel ensemble of neural networks using information content

Over the last decades, there has been a growing interest in applying artificial intelligence techniques to solve a spectrum of financial problems. A number of studies have shown promising results in using artificial neural networks (ANNs) to guide investment trading. Given the expanding role of ANNs in financial trading, this paper proposes the use of a hybrid neural network, which consists of two independent ANN architectures, and comparatively evaluates its performance against independent ANNs and econometric models in the trading of a financial‐engineered (synthetic) derivative composed of options on foreign exchange futures. We examine the financial profitability and the market timing ability of the competing neural network models and statistically compare their attributes with those based on linear and nonlinear statistical projections. A random walk model and the option pricing method are also included as benchmarks for comparison. Our empirical investigation finds that, for each of the currencies analysed, trading strategies guided by the proposed dual network are financially profitable and yield a more stable stream of investment returns than the other models. Statistical results strengthen the notion that diffusion of information contents and cross‐validation between the independent components within the dual network are able to reduce bias and extreme decision making over the long run. Moreover, the results are robust with respect to different levels of transaction costs. Copyright © 2009 John Wiley & Sons, Ltd.     

Modelling small‐business credit scoring by using logistic regression,neural networks and decision trees

Previous research on credit scoring that used statistical and intelligent methods was mostly focused on commercial and consumer lending. The main purpose of this paper is to extract important features for credit scoring in small‐business lending on a dataset with specific transitional economic conditions using a relatively small dataset. To do this, we compare the accuracy of the best models extracted by different methodologies, such as logistic regression, neural networks (NNs), and CART decision trees. Four different NN algorithms are tested, including backpropagation, radial basis function network, probabilistic and learning vector quantization, by using the forward nonlinear variable selection strategy. Although the test of differences in proportion and McNemar's test do not show a statistically significant difference in the models tested, the probabilistic NN model produces the highest hit rate and the lowest type I error. According to the measures of association, the best NN model also shows the highest degree of association with the data, and it yields the lowest total relative cost of misclassification for all scenarios examined. The best model extracts a set of important features for small‐business credit scoring for the observed sample, emphasizing credit programme characteristics, as well as entrepreneur's personal and business characteristics as the most important ones. Copyright © 2005 John Wiley & Sons, Ltd.     

消费信贷与家庭人力资本投资

已有丰富的文献考察了消费信贷对家庭消费和投资行为的影响，但少有研究关注其对家庭人力资本投资的影响。家庭人力资本投资对于人力资本积累、经济高质量发展至关重要。基于此，本文考察了信用卡使用对家庭人力资本投资的影响及其影响机制，主要发现如下：信用卡使用显著增加了家庭人力资本投资，且这一效应具有长期动态影响并对城市、高收入、高教育程度家庭影响更为显著，这意味着信用卡消费信贷可能会增加人力资本不平等。进一步研究发现，家庭会增加劳动力供给来应对人力资本支出的增加。机制分析表明，信用卡使用主要通过增加家庭消费投资、促进消费升级、缓解家庭预算约束三种途径促进家庭人力资本投资。在风险可控的前提下，引导消费信贷流向有利于实体经济发展的领域，可助力于消费升级与人力资本积累，从而为经济发展提供新动能。