Default Risk Prediction Research Articles

Comparing Performance of Machine Learning Algorithms for Default Risk Prediction in Peer to Peer Lending

The purpose of this research is to evaluate several popular machine learning algorithms for credit scoring for peer to peer lending. The dataset to fit the models is extracted from the official site of Lending Club. Several models have been implemented, including single classifiers (logistic regression, decision tree, multilayer perceptron), homogeneous ensembles (XGBoost, GBM, Random Forest) and heterogeneous ensemble classifiers like Stacked Ensembles. Results show that ensemble classifiers outperform single ones with Stacked Ensemble and XGBoost being the leaders.

A DBM-Based Ensemble Method for Improving Default Risk Prediction of Peer-to-Peer (P2P) Lending

With the peer-to-peer lending (P2P) business growing up, the most important influencing factor for the healthy development of this industry is the default risk of borrowers. Because the behavior between lenders and borrowers is real time, naturally large amounts of transaction data are being generated all the time. However, it is difficult to extract useful representative features and choose an appropriate model to predict the default risk of the borrowing behavior. In this paper, we proposed a (Deep Boltzmann Machines) DBM-based ensemble method for the default risk prediction in p2p lending, which is based on the real data generated by Lending Club company. Experimental results on the real world data indicate that our model is more effective and powerful with a 0.9093 explanation power.

基于机器学习方法的企业贷款违约风险预测

基于机器学习方法的企业贷款违约风险预测

The Prediction Analysis of Peer-to-Peer Lending Platforms Default Risk Based on Comparative Models

This paper examines the determinants of platform default risk using machine learning methods, including comprehensive models, and thus compares these models’ predictive abilities. To test platform’s default risk, this paper constructs three types of variables, which reflect a platform’s operating characteristics, customer feedback, and compliance capability. We find that the abnormal return tends to trigger default risk significantly. However, the default risk can be minimized if a platform has positive recommendations from customers and more transparent information disclosure or is affiliated as the member of the National Internet Finance Association of China. Empirical results indicate that the CART model outperforms the Random Forests model and Logit regression in predicting platform default risk. Our study sheds light on default risk prediction and thus can improve the government regulation ability.

Two-stage adaptive integration of multi-source heterogeneous data based on an improved random subspace and prediction of default risk of microcredit

Some scholars have shown that the machine learning methods based on a single-source data can successfully monitor the risks of formal financial activities, but not those of informal financial activities. This is because the data generated by formal financial activities, whether it is the structured or unstructured data, are of high quality and quantity, while the data generated by informal financial activities are not. Therefore, multi-source data are the key to monitor the risks of informal financial activities through machine learning. Although a few studies attempted to use multi-source data for financial risk prediction, they simply stack the obtained multi-source data, but ignore the original sources, heterogeneity, mutual redundancy and other characteristics of the data, so that the improvement of the prediction effect is not obvious. Therefore, TSAIB_RS method based on the two-stage adaptive integration of multi-source heterogeneous data was constructed in the paper, in which the data with different sources and different distributions were adaptively integrated. In order to test the reliability of TSAIB_RS method, the paper takes the default risk of microcredit in China as the test target and compares the prediction results of various test methods. It concludes that TSAIB_RS method can significantly improve the prediction effects.

An alternative statistical framework for credit default prediction

The purpose of this study is to introduce a gradient-boosting model that is robust to high-dimensional data and can produce a strong classifier by combining the predictors of many weak classifiers for credit default risk prediction. Therefore, this method is recommended for practical applications. This study compares the gradient-boosting model with four other well-known classifiers, namely, a classification and regression tree (CART), logistic regression (LR), multivariate adaptive regression splines (MARS) and a random forest (RF). Six real-world credit data sets are used for model validation. The performance of each model is compared using six performance measures, and a receiver operating characteristics (ROC) curve is plotted for the best classifiers of each data set. The empirical findings confirm that the gradient-boosting model is reliable and efficient across all of the performance criteria. In addition, LR and CART exhibit superior performances. The contributions of this study have theoretical and practical implications, as credit default risk prediction is a complicated and always contemporary issue.

Prediction of Default Risk in Peer-to-Peer Lending Using Structured and Unstructured Data

Using data from Lending Club, we analyzed funded loans between 2012 and 2013, the default status of which were mostly known in 2018. Our results showed that both the borrower characteristics and the conditions of the loan were significantly associated with the loan default rate. Results also showed that the sentiment of a user-written loan description influenced the borrower's loan interest rates. It contributes to expanding the scope of peer-to-peer (P2P) loan research by implementing unstructured data as a new model variable. Financial counselors need to consider the growth potential of the P2P loan market using data analysis: This will reveal niche market opportunities, enabling the development of services necessary for the safe supply of small loans at reasonable interest rates.

Listening to the investors: A novel framework for online lending default prediction using deep learning neural networks

Online peer-to-peer (P2P) lending has developed dramatically over the last decade in China. But this rapid boom carries potential risks. Investors have incurred incalculable losses due to the recent increase in fraudulent and/or unreliable online P2P platforms. Hence, predicting and identifying potential default risk platforms is crucial at this juncture. To achieve this end, we propose a two-step method which employs a deep learning neural network to extract keywords from investor comments and then utilizes a bidirectional long short-term memory (BiLSTM) based model to predict the default risk of platforms. Experimental results on real-world datasets of about 1000 platforms show that in the keyword extraction phase, our model can better capture semantic features from highly colloquial comment-text and achieve significant improvement over other baselines. Additionally, in the default platform prediction stage, our model achieves an F1 value of 80.34% in identifying potential problem platforms, outperforming four baselines by 23.37%, 5.71%, 8.93%, and 4.98% of improvement and comprehensively verifying the effectiveness of our method. Our study provides an alternative solution for platform default risk prediction issues and validates the effectiveness of investor comments in revealing the risk situation of online lending platforms.

Reconstructing China's Credit Default Risk Prediction Model: Based on the Perspective of Industry Sensitivity

Reconstructing China's Credit Default Risk Prediction Model: Based on the Perspective of Industry Sensitivity

Research on financial default risk prediction method based on big data model

Research on financial default risk prediction method based on big data model

Heterogeneous ensemble learning with feature engineering for default prediction in peer-to-peer lending in China

In recent years, peer-to-peer (P2P) lending in China, which is a new form of unsecured financing that uses the Internet, has boomed, but the consequent credit risk problems are inevitable. A key challenge facing P2P lending platforms is accurately predicting the default probability of the borrower of each loan using the default prediction model, which effectively helps the P2P lending platform avoid credit risks. The traditional default prediction model based on machine learning and statistical learning does not meet the needs of P2P lending platforms in terms of default risk prediction because for data-driven P2P lending, credit data have a large number of missing values, are high-dimensional and have class-imbalanced problems, which makes it difficult to effectively train the default risk prediction model. To solve the above problems, this paper proposes a new default risk prediction model based on heterogeneous ensemble learning. Three individual classifiers, extreme gradient boosting (XGBoost), a deep neural network (DNN) and logistic regression (LR), are used simultaneously with a liner weight ensemble strategy. In particular, this model is able to process missing values. After generating discrete and rank features, this model adds missing values to the model for self-training. Then, the hyperparameters are optimized by the XGBoost model to improve the performance of the prediction model. Finally, compared with the benchmark model, the proposed method significantly improves the accuracy of the prediction results. In conclusion, the prediction method proposed in this paper solves the class-imbalanced problem.

Transfer learning-based default prediction model for consumer credit in China

Financial institutions in China, such as banks, are encountering competitive impacts from Internet financial businesses. To address these impacts, financial institutions are seeking business innovations, such as an automatic credit evaluation system that is based on machine learning. Abundant new credit data are required in the implementation of new businesses to establish related risk evaluation models; however, new businesses lack data. Based on these insights, this paper innovatively proposes the idea of transfer learning, determines the similarity between traditional businesses and new businesses and transfers the data of traditional bank businesses to new business data to construct new training sets and to train small data sets. The reconstructed training data sets are used to train default risk prediction models, compare them with the benchmark models in the tests and validate the performance and adaptation of the default prediction model based on transfer learning technique. Our study highlights the commercial value of the transfer learning concept in the financial risk field and provides practitioners and management personnel with a decision basis.

Heterogeneous Ensemble for Default Prediction of Peer-to-Peer Lending in China

As a novel financing method, peer-to-peer (P2P) lending has drawn extensive attention as it provides those financers who cannot participate in the traditional financial market with funds. In P2P lending marketplaces, one of the crucial challenges that P2P online lending platforms are facing is to accurately predict the default risk of each loan by tapping into default prediction models, thus effectively helping P2P lending companies avoid credit risks. That traditional credit risk prediction models fail to meet the demand of P2P lending companies for default risk prediction, which is because of the uneven distribution of credit data samples in the P2P lending marketplaces (i.e., the default sampled data are scarce). In this paper, we designed a multi-round ensemble learning model based on heterogeneous ensemble frameworks to predict default risk. In this model, an extreme gradient boosting (XGBoost) is initially used for ensemble learning, and the XGBoost, deep neural network, and logistic regression are then regarded as heterogeneous individual learners to undergo a linear weighted fusion. To verify the designed default risk prediction model, real credit data from a famous P2P online lending marketplace in China were used in a test. The results of the experiment indicate that this model can effectively increase the predictive accuracy compared with traditional machine learning models and ensemble learning models.

Parametric prediction on default risk of Chinese listed tourism companies by using random oversampling, isomap, and locally linear embeddings on imbalanced samples

This research pioneers the default risk parametric prediction of Chinese tourism companies with random oversampling and manifold learning for parametric modelling on imbalanced samples to relax the requirement on sample availability. Four specific approaches were employed: standardization; standardization→random oversampling; standardization→isomap+locally linear embeddings; and standardization→random oversampling→isomap+locally linear embeddings. Empirical results indicate that: random oversampling successfully improved the tourism default risk prediction; the integration of isomap and locally linear embeddings is beneficial in default risk prediction using highly skewed tourism data with absolute minority samples; and after the use of random oversampling on initial data, the integrated approach improved in forecasting tourism default risk prior to two years versus one year.

An enforced support vector machine model for construction contractor default prediction

Abstract The financial health of construction contractors is critical in successfully completing a project, and thus default prediction is highly concerned by owners and other stakeholders. In other industries many previous studies employ support vector machine (SVM) or other Artificial Neural Networks (ANN) methods for corporate default prediction using the sample-matching method, which produces sample selection biases. In order to avoid the sample selection biases, this paper used all available firm-years samples during the sample period. Yet this brings a new challenge: the number of non-defaulted samples greatly exceeds the defaulted samples, which is referred to as between-class imbalance. Although the SVM algorithm is a powerful learning process, it cannot always be applied to data with extreme distribution characteristics. This paper proposes an enforced support vector machine-based model (ESVM model) for the default prediction in the construction industry, using all available firm-years data in our sample period to solve the between-class imbalance. The traditional logistic regression model is provided as a benchmark to evaluate the forecasting ability of the ESVM model. All financial variables related to the prediction of contractor default risk as well as 7 variables selected by the Multivariate Discriminant Analysis (MDA) stepwise method are put in the models for comparison. The empirical results of this paper show that the ESVM model always outperforms the logistic regression model, and is more convenient to use because it is relatively independent of the selection of variables. Thus, we recommend the proposed ESVM model as an alternative to the traditionally used logistic model.

Prediction of default risk: An options-based approach applied to the Brazilian banking sector

This article proposes a methodological framework to construct an early warning system for the Banking sector. It employs an options-based methodology to estimate default risk for six major Brazilian banks and shows that these measures have informational content. In addition, the options-based indicator is compared with market-based financial fragility indicators. Results show that these indicators are useful for risk managers and regulators, especially during crisis. Furthermore, option-based methods are preferable to classify banks. Finally, there is some evidence of lack of market discipline in the early 1990s.

Macroeconomic Growth, Real Estate Market Conditions, and the Time Series Dynamics of CMBS Loan Default Risk

Recent financial market turmoil calls for better understanding of the default risk of mortgages and mortgage-related securities. In this paper, I study commercial mortgage default risk using the performance data of nearly 50,000 CMBS loans. Particularly, I examine the time series dynamics of commercial mortgage default risk and the underlying systematic risk factors in both the macroeconomy and the real estate market. A first-passage model with equilibrium macroeconomic dynamics is presented, and the default hazard rate is solved. The solutions are then put into a state space form and the model is estimated with real world mortgage performance data using extended Kalman filter. Results show large variations of default risk over time in the commercial mortgage market, and that these variations are well explained by two mean-reverting latent risk factors -- a macroeconomic factor and a commercial property market-specific factor. The model and the results can be used in default risk prediction, hedging and pricing.

Macroeconomic Conditions, Systematic Risk Factors, and the Time Series Dynamics of Commercial Mortgage Credit Risk

I study the time series dynamics of commercial mortgage credit risk and the unobservable systematic risk factors underlying those dynamics. A first-passage model with equilibrium macroeconomic dynamics is presented, and the default hazard rate is solved. The solutions are then put into a state space form and estimated with real world commercial mortgage performance data using extended Kalman filter. Results show large variations of credit risk over time in the commercial mortgage market, and that these variations are well explained by two mean-reverting latent risk factors. One is the macroeconomic factor and the other is a commercial property market-specific factor. The model and the results will be useful in default risk prediction, hedging and pricing.

Stochastic modelling and prediction of contractor default risk

The purpose of this paper is to describe a systematic framework of stochastic modelling and prediction of financial default risk of construction contractors. Net-worth-to-asset ratio is identified as an index for default process modelling. The default condition is defined as when the ratio becomes negative the first time. A mean-reverting dynamic model for the contractor default process is found by statistical analysis and is justified by using the theory of optimal capital structure. The stochastic modelling of default uses the time to default as the fundamental random variable. A discrete time trinomial Markov chain model is developed to assess default risk in terms of a cumulative default probability function, a default probability function, and the mean and variance of time to default. Practical examples are given to illustrate the stochastic methods. A default discriminant study on a group of contractors and publicly traded companies validates the methods, and indicates a high predictability of events of default and declines of credit rating.