Narrow Results

While feedforward neural networks have been widely accepted as effective tools for solving classification problems, the issue of finding the best network architecture remains unresolved, particularly so in real-world problem settings. We address this issue in the context of credit card screening, where it is important to not only find a neural network with good predictive performance but also one that facilitates a clear explanation of how it produces its predictions. We show that minimal neural networks with as few as one hidden unit provide good predictive accuracy, while having the added advantage of making it easier to generate concise and comprehensible classification rules for the user. To further reduce model size, a novel approach is suggested in which network connections from the input units to this hidden unit are removed by a very straightaway pruning procedure. In terms of predictive accuracy, both the minimized neural networks and the rule sets generated from them are shown to compare favorably with other neural network based classifiers. The rules generated from the minimized neural networks are concise and thus easier to validate in a real-life setting.

Numerous studies have contributed to efforts to boost the accuracy of the credit scoring model. Especially interesting are recent studies which have successfully developed the hybrid approach, which advances classification accuracy by combining different machine learning techniques. However, to achieve better credit decisions, it is not enough merely to increase the accuracy of the credit scoring model. It is necessary to conduct meaningful supplementary analyses in order to obtain knowledge of causal relations, particularly in terms of significant conceptual patterns or structures involving attributes used in the credit scoring model. This paper proposes a solution of integrating data preprocessing strategies and the Bayesian network classifier with the tree augmented Na"ıve Bayes search algorithm, in order to improve classification accuracy and to obtain improved knowledge of causal patterns, thus enhancing the validity of credit decisions.

The assessment of the risk of default on credit is important for financial institutions. Different Artificial Neural Networks (ANN) have been suggested to tackle the credit scoring problem, however, the obtained error rates are often high. In the search for the best ANN algorithm for credit scoring, this paper contributes with the application of an ANN Training Algorithm inspired by the neurons' biological property of metaplasticity. This algorithm is especially efficient when few patterns of a class are available, or when information inherent to low probability events is crucial for a successful application, as weight updating is overemphasized in the less frequent activations than in the more frequent ones. Two well-known and readily available such as: Australia and German data sets has been used to test the algorithm. The results obtained by AMMLP shown have been superior to state-of-the-art classification algorithms in credit scoring.

It is crucial to correctly assess whether a potential borrower can repay the loan in the credit scoring model. The credit loan data has a serious data imbalance because the number of defaulters is far less than the nondefaulters. However, most current methods for dealing with data imbalance are designed to improve the classification performance of minority data, which will reduce the performance of majority data. For a financial institution, the economic loss caused by the decrease in the classification performance of nondefaulters (majority data) cannot be ignored. This paper proposes an ensemble learning method based on one-class and binary classification (EMOBC) for credit scoring. The purpose is to improve the classification accuracy of the minority class while mitigating the loss of classification accuracy of the majority class as much as possible. EMOBC uses undersampling for the majority class (nondefault samples in credit scoring) and perform binary-class learning on the balanced data to improve the classification accuracy of the minority. To alleviate the decline in classification performance of the majority class, EMOBC uses one-class and binary collaborative classification to train classifiers. The classification result is determined by the average of one-class and binary-class classifiers. The experimental results show that EMOBC has good comprehensive performance compared with the existing methods.

Nowadays, one of the biggest challenges in banking sector, certainly, is assessment of the client’s creditworthiness. In order to improve the decision-making process and risk management, banks resort to using data mining techniques for hidden patterns recognition within a wide data. The main objective of this study is to build a high-performance customized credit scoring model. The model named Reliable client is based on Bank’s real dataset and originally built by applying four different classification algorithms: decision tree (DT), naive Bayes (NB), generalized linear model (GLM) and support vector machine (SVM). Since it showed the greatest results, but also seemed as the most appropriate algorithm, the adopted model is based on GLM algorithm. The results of this model are presented based on many performance measures that showed great predictive confidence and accuracy, but we also demonstrated significant impact of data pre-processing on model performance. Statistical analysis of the model identified the most significant parameters on the model outcome. In the end, created credit scoring model was evaluated using another set of real data of the same Bank.

Credit scoring models are very important tools for financial institutions to make credit granting decisions. In the last few decades, many quantitative methods have been used for the development of credit scoring models with focus on maximizing classification accuracy. This paper proposes the credit scoring models with the area under receiver operating characteristics curve (AUC) maximization based on the new emerged support vector machines (SVM) techniques. Three main SVM models with different features weighted strategies are discussed. The weighted SVM credit scoring models are tested using 10-fold cross validation with two real world data sets and the experimental results are compared with other six traditional methods including linear regression, logistic regression, k nearest neighbor, decision tree, and neural network. Results demonstrate that weighted 2-norm SVM with radial basis function (RBF) kernel function and t-test feature weighting strategy has the overall better performance with very narrow margin than other SVM models. However, it also consumes more computational time. In considering the balance of performance and time, least squares support vector machines (LSSVM) with RBF kernel maybe a better choice for large scale credit scoring applications.

Consumer credit scoring is considered as a crucial issue in the credit industry. SVM has been successfully utilized for classification in many areas including credit scoring. Kernel function is vital when applying SVM to classification problem for enhancing the prediction performance. Currently, most of kernel functions used in SVM are single kernel functions such as the radial basis function (RBF) which has been widely used. On the basis of the existing kernel functions, this paper proposes a multi-kernel function to improve the learning and generalization ability of SVM by integrating several single kernel functions. Chaos particle swarm optimization (CPSO) which is a kind of improved PSO algorithm is utilized to optimize parameters and to select features simultaneously. Two UCI credit data sets are used as the experimental data to evaluate the classification performance of the proposed method.

In this paper, we present an approach for sample selection using an ensemble of neural networks for credit scoring. The ensemble determines samples that can be considered outliers by checking the classification accuracy of the neural networks on the original training data samples. Those samples that are consistently misclassified by the neural networks in the ensemble are removed from the training dataset. The remaining data samples are then used to train and prune another neural network for rule extraction. Our experimental results on publicly available benchmark credit scoring datasets show that by eliminating the outliers, we obtain neural networks with higher predictive accuracy and simpler in structure compared to the networks that are trained with the original dataset. A rule extraction algorithm is applied to generate comprehensible rules from the neural networks. The extracted rules are more concise than the rules generated from networks that have been trained using the original datasets.

Recent interest in financial technology (fintech) lending business has caused increasing challenges of credit scoring models using bank account activity information. Our work aims to develop a new credit scoring method based on bank account activity information. This method incorporates borrower firms’ segment-level heterogeneity, such as a segment of sales size and firm age. We employ Bayesian hierarchical modeling, which mitigates data sparsity issue due to data segmentation. We describe our modeling procedures, including data handling and variable selection. Empirical results show that our model outperforms the traditional logistic model for credit scoring in information criterion. Our model realizes advanced credit scoring based on bank account activity information in fintech lending businesses, taking segment-specific features into credit risk assessment.

Credit scoring is used by institutions in managing the credit risk associated with their borrowers. In addition to the traditional variables used in credit scoring, recent developments in machine learning have allowed the inclusion of data sources, such as ongoing payment behaviors and patterns, creating more robust and dynamic models. While such studies have more commonly been done for credit card data, there is limited research in the context of microfinance. Investigating this niche, this study focuses on uncovering how delinquency can be characterized and whether delinquency states exist within microfinance loans using a dataset from a microfinance institution (MFI) in the Philippines. Further, Markov chains and transition state matrices from the dataset were used to predict repayment sequences, and factors such as borrower characteristics, loan information and repayment behaviors were used in predicting the final outcome of the loans. The findings show that Markov chains can predict repayment sequences using a transition state matrix based on the full term of the loan. This model outperformed other models created from transition state matrices based on the first 8, 12, and 17 weeks of the loan term, indicating the repayment performance in the first weeks is not representative of the full term and cannot be used to accurately predict future repayment patterns. The equivalent second-order Markov chains were also examined which resulted in better predictions generally. Meanwhile, in predicting the final outcome of the loans, random forest models outperformed decision tree models using an accuracy and modified accuracy metric for evaluation. Across all models assessed in this study, behavioral characteristics and payment patterns consistently fared higher in terms of feature importance than borrower and loan characteristics. Particular to tiering delinquency states, the combination of average days overdue and delinquency streak appeared to be fitting resulting in delinquency tiers Onset Delinquency and Significant Delinquency. The ability to predict repayments aims to provide MFIs with better oversight of their loan portfolios. With an emphasis on the early detection of delinquency, predictive models using relevant features, such as those identified in this study, may allow the implementation mitigating actions at specific time periods to improve overall repayment rates among borrowers of MFIs.

Credit scoring prediction technique is to assist the lender in making decisions. Accurately predicting user’s credit scoring is a challenging issue. In this chapter, a new credit scoring prediction method is proposed, in which the users and their relationships are modeled as a weighted undirected graph, and then the credit scoring prediction problem is reduced to a graph signal reconstruction problem. The proposed method utilizes both the information of the unlabeled data and the location information of the samples in the feature space, and thus achieves an excellent predictive performance. Experimental results on the open UCI German credit dataset are compared with those of seven classical models that the prediction performance of the proposed model is significantly better than that of the reference models. The Friedman test indicates that the experimental results have statistical significance.

Credit card lending in Asia has grown rapidly since the 1997 Asian Financial Crisis. But this growth so far has seen several episodes of sharp booms and busts, posing new risks to financial stability. This chapter attempts to learn more about the credit card markets and shed light on and draw lessons from some of the core common elements in the three recent episodes of credit card lending distress in Asia. Of the episodes in Hong Kong, Korea and Taiwan during the 2000s, Korea experienced the most serious boom-bust cycle in 2003. Therefore, we take a closer look at the credit card lending crisis there. Easy monetary conditions, financial liberalization, economies of scales in the credit card business, limited information reporting and sharing, various forms of the principal-agent problem, and the seasoning effect of a fast-growing lending portfolio all contributed to unsustainable credit lending boom. Eventually, excessive indebtedness led to rising credit cost and tighter lending standard, contracting credit, further credit loss, big balance sheet adjustments, risks to the financial system and damages to the economy. Policymakers in Asian emerging markets such as China and India need to learn about the risks arising from this type of consumer lending and respond with enhanced supervisory capacity, better market infrastructure, and appropriate prudential measures.

The following sections are included:

• Introduction
• Credit Scoring in Microfinance Institutions: The Literature
• Data and Model
• Empirical Results
• Conclusion
• Notes
• References