Credit Risk Classification Research Articles

Credit Risk Classification and Prediction Based on Deep Neural Network Algorithm

Credit Risk Classification and Prediction Based on Deep Neural Network Algorithm

Credit Risk Classification Prediction Based on Optimised Adaboost Algorithm with Long Short-Term Memory Neural Network (LSTM)

Credit Risk Classification Prediction Based on Optimised Adaboost Algorithm with Long Short-Term Memory Neural Network (LSTM)

A SC Financial Credit Risk Assessment Model Based on Particle Filter and SVM with Gain Information

The accuracy of credit risk prediction in SC financing is critical for many enterprises, based on machine learning algorithms can be good for SME credit risk assessment research, for this reason, this paper establishes a combinatorial model that can improve credit risk prediction, using support vector machine (SVM) and particle filtering to achieve credit risk classification and prediction, we and introduce information gain (IG) to extract the prediction of The model uses SVM and particle filtering to classify and predict credit risk, and we introduce information gain (IG) to extract feature variables that contribute significantly to the prediction results and optimize model feature inputs. Compared with the benchmark model, the prediction accuracy of the model in this paper is 97.62%, which is 8.97% higher than that of SVM, and the performance of IG with feature optimization improves the prediction accuracy by another 3%.

Semi-supervised heterogeneous domain adaptation for few-sample credit risk classification

Credit risk classification is a crucial area in machine learning-enhanced financial decision support systems, and numerous studies have achieved significant progress. However, data in the modern financial landscape is inherently complex, characterized by a lack of labeled data, cross-domain heterogeneity, and class imbalance. These three major problems hinder the achievement of credit risk classification. Therefore, we propose a semi-supervised heterogeneous domain adaptation method and an imbalanced data augmentation method to overcome these challenges with a single neural network-based model. Experimental results obtained from four representative benchmark datasets confirm the superior performance of the proposed method.

Enhancing banking governance: A machine learning-based credit risk classification

<p>Risk management in the banking sector has gained heightened significance following the 2008 Global Financial Crisis. With the advent of Machine Learning (ML) techniques, financial institutions are increasingly turning to Artificial Intelligence (AI) for enhanced risk assessment and management. This paper introduces a systematic protocol for implementing a decision tree classifier tailored for credit risk classification. Additionally, we develop a user-friendly web application utilizing the Flask framework and Python Pickle library. This application offers customers an intuitive interface to input their attributes and receive predictions regarding their credit risk classification. Our empirical findings demonstrate that the Support Vector Machine (SVM) achieves a commendable accuracy of 77% in classifying customers based on their banking data. Furthermore, the web application proves to be an effective means for customers to interact with the ML model, enhancing accessibility and user engagement. These outcomes underscore the substantial benefits that ML techniques can bring to the banking industry, enabling improved risk detection and management while concurrently enhancing customer service delivery.</p>

Enhancing Supervised Model Performance in Credit Risk Classification Using Sampling Strategies and Feature Ranking

For the financial health of lenders and institutions, one important risk assessment called credit risk is about correctly deciding whether or not a borrower will fail to repay a loan. It not only helps in the approval or denial of loan applications but also aids in managing the non-performing loan (NPL) trend. In this study, a dataset provided by the LendingClub company based in San Francisco, CA, USA, from 2007 to 2020 consisting of 2,925,492 records and 141 attributes was experimented with. The loan status was categorized as “Good” or “Risk”. To yield highly effective results of credit risk prediction, experiments on credit risk prediction were performed using three widely adopted supervised machine learning techniques: logistic regression, random forest, and gradient boosting. In addition, to solve the imbalanced data problem, three sampling algorithms, including under-sampling, over-sampling, and combined sampling, were employed. The results show that the gradient boosting technique achieves nearly perfect Accuracy, Precision, Recall, and F1score values, which are better than 99.92%, but its MCC values are greater than 99.77%. Three imbalanced data handling approaches can enhance the model performance of models trained by three algorithms. Moreover, the experiment of reducing the number of features based on mutual information calculation revealed slightly decreasing performance for 50 data features with Accuracy values greater than 99.86%. For 25 data features, which is the smallest size, the random forest supervised model yielded 99.15% Accuracy. Both sampling strategies and feature selection help to improve the supervised model for accurately predicting credit risk, which may be beneficial in the lending business.

A shapelet-based behavioral pattern extraction method for credit risk classification with behavior sparsity

In this paper, a shapelet-based behavioral pattern extraction method (Behavior2Shapelets) is proposed to address the behavior-sparsity issue in credit risk classification. In the proposed method, three steps are involved. First, by introducing the shapelets module, the proposed model can learn the behavioral patterns from the sparse behavioral features. Then, these extracted patterns are combined with a logistic regression model to construct an interpretable classifier. Finally, to enhance the robustness of the model, a dynamic strategy for threshold determination is used, which is based on the Kolmogorov-Smirnov statistic. The performance of the Behavior2Shapelets model is validated by using the Taiwan credit dataset and its seven derivative datasets. The empirical results demonstrate that the Behavior2Shapelets model outperforms the baseline methods in terms of classification accuracy and robustness. This highlights the feasibility of the proposed model as a solution to the behavior-sparsity problem in credit risk classification.

IMPROVED SUPPORT VECTOR MACHINE PERFORMANCE USING PARTICLE SWARM OPTIMIZATION IN CREDIT RISK CLASSIFICATION

In Classification using Support Vector Machine (SVM), each kernel has parameters that affect the classification accuracy results. This study examines the improvement of SVM performance by selecting parameters using Particle Swarm Optimization (PSO) on credit risk classification, the results of which are compared with SVM with random parameter selection. The classification performance is evaluated by applying the SVM classification to the Credit German benchmark credit data set and the private credit data set which is a credit data set issued from a local bank in North Sumatra. Although it requires a longer execution time to achieve optimal accuracy values, the SVM+PSO combination is quite effective and more systematic than trial and error techniques in finding SVM parameter values, so as to produce better accuracy. In general, the test results show that the RBF kernel is able to produce higher accuracy and f1-scores than linear and polynomial kernels. SVM classification with optimization using PSO can produce better accuracy than classification using SVM without optimization, namely the determination of parameters randomly. Credit data classification accuracy increased to 92.31%.

Forecasting financial markets and credit risk classification using genetic folding algorithm

Making choices is a basic managerial skill. Some financial markets employ credit monitoring and risk analysis to control risk, forcing financial services and consultancy businesses to develop quantitative decision-making models. Incorporating several methods into the classification system for credit risk leads to more substantial research. Several quantitative credit scoring methods are now available for credit risk assessment. Some disadvantages of metaheuristic techniques include local maxima and early convergence. As a consequence, GFLibPy, a free and open source genetic folding (GF) Python toolkit, is provided. This paper aims to provide: 1) a technical explanation of the 'state-of-the-art' GF algorithm in the credit approval and German credit databases; 2) a road map to help management adopt GFLibPy technologies in banking; 3) GFLibPy increases classification accuracy for new loan applications. In testing phase, the GFLibPy performance varied from 81% to 93%.

Forecasting financial markets and credit risk classification using genetic folding algorithm

Forecasting financial markets and credit risk classification using genetic folding algorithm

A Gaussian process‐based approach toward credit risk modeling using stationary activations

AbstractThe task of predicting the risk of defaulting of a lender using tools in the domain of AI is an emerging one and in growing demand, given the revolutionary potential of AI. Various attributes like income, properties acquired, educational status, and many other socioeconomic factors can be used to train a model to predict the possibilities of nonrepayment of a loan or its chances. Most of the techniques and algorithms used in this regard previously do not submit any attention to the uncertainty in predictions for out of distribution (OOD) in a dataset, which contributes to overfitting, leading to relatively lower accuracy for predicting these data points. Specifically, for credit risk classification, this is a serious concern, given the structure of the available datasets and the trend they follow. With a focus on this issue, we propose a robust and better methodology that uses a recent and efficient family of nonlinear neural network activation functions, which mimics the properties induced by the widely‐used Matérn family of kernels in Gaussian process (GP) models. We tested the classification performance metrics on three openly available datasets after prior preprocessing. We achieved a high mean classification accuracy of 87.4% and a lower mean negative log predictive density loss of 0.405.

Credit risk classification: an integrated predictive accuracy algorithm using artificial and deep neural networks

Credit risk classification: an integrated predictive accuracy algorithm using artificial and deep neural networks

Evaluation of SMEs’ Credit Decision Based on Support Vector Machine-Logistics Regression

This article uses support vector machines, logistics regression, and other methods for the comprehensive evaluation of credit decision-making of small, medium, and microenterprises and comprehensively uses software programming such as MATLAB and SPSS Modeler to solve the problem. The results, such as credit risk evaluation index system, credit risk classification model, and credit decision-making comprehensive evaluation model, are obtained. Finally, this article starts from the credit decision of small, medium, and microenterprises and provides theoretical and practical suggestions for banks to control the risks of small, medium, and microenterprises and their own development.

Missing Data Preprocessing in Credit Classification: One-Hot Encoding or Imputation?

ABSTRACT Missing data has become an increasingly serious problem in credit risk classification. A one-hot encoding-based data preprocessing method is proposed to solve the missing data problem in credit classification. In this paradigm, the proposed missing-data preprocessing method is first used to deal with missing values to fill in the incomplete dataset. Then the classification and regression tree (CART) model is applied on the completed dataset to measure performances of different preprocessing methods. The experimental results indicate that the proposed one-hot encoding method performs the best when the missing rate is high. When missing rate is low, random sample (RS) imputation method performs better though it entails a greater computational cost than other imputation methods listed in this study. In particular, for high-missing-rate coupled with data-imbalance issue, the proposed one-hot encoding based imputation method shows not only high accuracy, but also great robustness and needs less of computational time.

Can machine learning paradigm improve attribute noise problem in credit risk classification?

In this paper, a dual-voting-based learning paradigm is proposed to solve attribute noise problem in credit risk classification. In the proposed learning paradigm, three stages are involved. In the first stage, four indexes are introduced to evaluate the noise level of attributes. In the second stage, attributes with different noise levels are divided into different attribute sets in accordance with the dual-voting results of noise level. In the third stage, credit datasets with different attributes sets are dealt with different learning strategies and different de-noising methods for comparison purpose. In the proposed learning paradigm, a classification and regression tree (CART) model is adopted as the generic classifier to evaluate the performance on training datasets generated by different learning strategies and noise reduction methods. In addition, the performance of all learning strategies on sparse data with attribute noise is also discussed. Experimental results show that the proposed learning paradigm performs better than the benchmarks to solve the attribute noise problem not only in accuracy and its stability, but also in speediness. Further analysis indicates that the sparse data with attribute noise can further improve the stability of accuracy for a specific de-noising method. This implies that the proposed dual voting-based learning paradigm is a promising solution to attribute noise reduction in credit risk classification.

Domain Adaptation Learning Based on Structural Similarity Weighted Mean Discrepancy for Credit Risk Classification

Domain adaptation learning is an effective method for leveraging knowledge from the source domain with labels to a target domain without any labels. However, most prior methods have neglected the contribution of each sample to the integral measure when adaptively matching either marginal distribution, conditional distribution, or both between domains. In this article, an improved algorithm based on mean discrepancy embedding with structural similarity is proposed, which aims at the contribution of each sample to the integral measure on the performance of target domain learning model by using labeled source samples and unlabeled target samples. The discrepancy between both marginal and conditional distribution are minimized with dimensionality reduction procedure to feature extraction with structural similarity weights for all samples from the source and target domains. The results of empirical analysis demonstrate that the proposed method has better performance over several state-of-the-art methods in credit risk classification.`

A VNS-EDA Algorithm-Based Feature Selection for Credit Risk Classification

Many quantitative credit scoring models have been developed for credit risk assessment. Irrelevant and redundant features may deteriorate the performance of credit risk classification. Feature selection with metaheuristic techniques can be applied to excavate the most significant features. However, metaheuristic techniques suffer from various issues such as being trapped in local optimum and premature convergence. Therefore, in this article, a hybrid variable neighborhood search and estimation of distribution technique with the elitist population strategy is proposed to identify the optimal feature subset. Variable neighborhood search with the elitist population strategy is used to direct its local searching in order to optimize the ergodicity, avoid premature convergence, and jump out of the local optimum in the searching process. The probabilistic model attempts to capture the probability distribution of the promising solutions which are biased towards the global optimum. The proposed technique has been tested on both publicly available credit datasets and a real-world credit dataset in China. Experimental analysis demonstrates that it outperforms existing techniques in large-scale credit datasets with high dimensionality, making it well suited for feature selection in credit risk classification.

Credit Risk Classification in Peer-to-Peer Marketplaces: The Nexus of Neural Network Approach

Financial innovation in recent years have prominently contributed to the growth of Peer-to-Peer lending marketplaces allowing individual and businesses to secure loans on a common internet-based network. Similar to the ‘bricks and mortar’ banking system, online lending is coupled with the problem of information asymmetry. Borrower risk assessment has henceforth become the major concerns of various platforms that aim to reducing information imbalance towards mitigating credit risk. In this article, authors compared two learning algorithms – Logistic regression and Artificial Neural Network to classify borrowers based on loan repayment schedule. We revealed that both approaches were robust in classifying late borrowers with logistic regression being 0.02% more robust than Neural Network. Regarding variable relative importance, gender is considered the least important variable whereas terms-of-repayment is the most important variable affecting borrowers’ intention to pay off loans. Even though our study contributes to existing literature, it is however not limited to determining factors that may affect lenders’ investment decision in social lending.

A DBN-based resampling SVM ensemble learning paradigm for credit classification with imbalanced data

Credit risk assessment is often accompanied with sampling data imbalance. For this reason, this paper tries to propose a deep belief network (DBN) based resampling support vector machine (SVM) ensemble learning paradigm to solve imbalanced data problem in credit classification. In this paradigm, a bagging algorithm is first used to generate variable training subsets to make the subsets rebalanced and suitable in size. Then the SVM model is used as individual base classifier to formulate diverse ensemble input members. Finally, the DBN model is applied as an ensemble method to fuse the input members to aggregate the classification results. In addition, the weights of different classes are changed by introducing a revenue matrix in terms of revenue-sensitive technique, which helps to make the results more reasonable. The experimental results indicate that the classification performance are improved effectively when the DBN-based ensemble strategy is integrated with re-sampling techniques, especially in imbalanced-data problem, implying that the proposed DBN-based resampling SVM ensemble learning paradigm can be used as a promising tool for credit risk classification with imbalanced data.

Neural Networks in Credit Risk Classification of Companies in the Construction Sector

The financial sector (banks, financial institutions, etc.) is the sector most exposed to financial and credit risk, as one of the basic objectives of banks' activity (as a specific enterprise) is granting credit and loans. Because credit risk is one of the problems constantly faced by banks, identification of potential good and bad customers is an extremely important task. This paper investigates the use of different structures of neural networks to support the preliminary credit risk decision-making process. The results are compared among the models and juxtaposed with real-world data. Moreover, different sets and subsets of entry data are analyzed to find the best input variables (financial ratios).