Armando Vieira

Improving bankruptcy prediction with Hidden Layer Learning Vector Quantization

Abstract A Hidden Layer Learning Vector Quantization (HLVQ), neural network-learning algorithm is used for correcting the outputs of Multilayer Perceptrons (MLP) for predicting corporate bankruptcy. We call this method HLVQ-C, and it is shown that it outperforms both discriminant analysis and traditional neural networks while significantly reducing type I error, which is the type of error that has the highest costs for banks. Moreover, our approach gives an estimation of the prediction robustness thus providing a useful measure of credit risk, which is of great interest for banks, insurance companies and creditors in general. We also show that unbalanced samples, containing more financially sound firms than bankrupt firms, place a strong bias on the classifiers thus leading to a deterioration of type I error accuracy. Although many studies have been published on bankruptcy prediction using neural networks or discriminant analysis, they used mainly US or UK samples of very limited size. Our study is based on industrial French firms, uses a data-set of 583 bankrupt firms over the period 1998–2000 and tests the effects of different proportions of non-bankrupt firms in the sample. Attention was also given to feature selection to reduce the dimensionality of the problem.

Clustering and visualization of bankruptcy trajectory using self-organizing map

Bankruptcy trajectory reflects the dynamic changes of financial situation of companies, and hence make possible to keep track of the evolution of companies and recognize the important trajectory patterns. This study aims at a compact visualization of the complex temporal behaviors in financial statements. We use self-organizing map (SOM) to analyze and visualize the financial situation of companies over several years through a two-step clustering process. Initially, the bankruptcy risk is characterized by a feature self-organizing map (FSOM), and therefore the temporal sequence is converted to the trajectory vector projected on the map. Afterwards, the trajectory self-organizing map (TSOM) clusters the trajectory vectors to a number of trajectory patterns. The proposed approach is applied to a large database of French companies spanning over four years. The experimental results demonstrate the promising functionality of SOM for bankruptcy trajectory clustering and visualization. From the viewpoint of decision support, the method might give experts insight into the patterns of bankrupt and healthy company development.

A genetic algorithm-based approach to cost-sensitive bankruptcy prediction

The prediction of bankruptcy is of significant importance with the present-day increase of bankrupt companies. In the practical applications, the cost of misclassification is worthy of consideration in the modeling in order to make accurate and desirable decisions. An effective prediction system requires the integration of the cost preference into the construction and optimization of prediction models. This paper presents an evolutionary approach for optimizing simultaneously the complexity and the weights of learning vector quantization network under the symmetric cost preference. Experimental evidences on a real-world data set demonstrate the proposed algorithm leads to significant reduction of features without the degradation of prediction capability.

Enhanced default risk models with SVM

Default risk models have lately raised a great interest due to the recent world economic crisis. In spite of many advanced techniques that have extensively been proposed, no comprehensive method incorporating a holistic perspective has hitherto been considered. Thus, the existing models for bankruptcy prediction lack the whole coverage of contextual knowledge which may prevent the decision makers such as investors and financial analysts to take the right decisions. Recently, SVM+ provides a formal way to incorporate additional information (not only training data) onto the learning models improving generalization. In financial settings examples of such non-financial (though relevant) information are marketing reports, competitors landscape, economic environment, customers screening, industry trends, etc. By exploiting additional information able to improve classical inductive learning we propose a prediction model where data is naturally separated into several structured groups clustered by the size and annual turnover of the firms. Experimental results in the setting of a heterogeneous data set of French companies demonstrated that the proposed default risk model showed better predictability performance than the baseline SVM and multi-task learning with SVM.

A training algorithm for classification of high-dimensional data

Abstract We propose an algorithm for training multi layer preceptrons (MLP) for classification problems, that we named hidden layer learning vector quantization. It consists of applying learning vector quantization to the last hidden layer of a MLP and it gave very successful results on problems containing a large number of correlated inputs. It was applied with excellent results on classification of Rutherford backscattering spectra and to a benchmark problem of image recognition.

Extracting discriminative features using non-negative matrix factorization in financial distress data

In the recent financial crisis the incidence of important cases of bankruptcy led to a growing interest in corporate bankruptcy prediction models. In addition to building appropriate financial distress prediction models, it is also of extreme importance to devise dimensionality reduction methods able to extract the most discriminative features. Here we show that Non-Negative Matrix Factorization (NMF) is a powerful technique for successful extraction of features in this financial setting. NMF is a technique that decomposes financial multivariate data into a few basis functions and encodings using non-negative constraints. We propose an approach that first performs proper initialization of NMF taking into account original data using K-means clustering. Second, builds a bankruptcy prediction model using the discriminative financial ratios extracted by NMF decomposition. Model predictive accuracies evaluated in real database of French companies with statuses belonging to two classes (healthy and distressed) are illustrated showing the effectiveness of our approach.

Learning manifolds for bankruptcy analysis

We apply manifold learning to a real data set of distressed and healthy companies for proper geometric tunning of similarity data points and visualization. While Isomap algorithm is often used in unsupervised learning our approach combines this algorithm with information of class labels for bankruptcy prediction. We compare prediction results with classifiers such as Support Vector Machines (SVM), Relevance Vector Machines (RVM) and the simple k-Nearest Neighbor (KNN) in the same data set and we show comparable accuracy of the proposed approach.

A stable credit rating model based on learning vector quantization

Credit rating is involved in many financial applications to estimate the creditworthiness of corporations or individuals. In addition to building accurate credit rating models, the stability of models is of significant importance to economic performance. In this work we propose a methodology based on learning vector quantization (LVQ) to develop a credit rating model. This model is applied to a French database of private companies over a period of several years. LVQ is trained and calibrated in a supervised way using data from 2006 and then applied to the remaining years. We analyze one year transition matrix and show that the model is capable to create robust and stable classes to rank companies.

Financial distress model prediction using SVM

Financial distress prediction is of great importance to all stakeholders in order to enable better decision-making in evaluating firms. In recent years, the rate of bankruptcy has risen and it is becoming harder to estimate as companies become more complex and the asymmetric information between banks and firms increases. Although a great variety of techniques have been applied along the years, no comprehensive method incorporating an holistic perspective had hitherto been considered. Recently, SVM+ a technique proposed by Vapnik [17] provides a formal way to incorporate privileged information onto the learning models improving generalization. By exploiting additional information to improve traditional inductive learning we propose a prediction model where data is naturally separated into several groups according to the size of the firm. Experimental results in the setting of a heterogeneous data set of French companies demonstrated that the proposed model showed superior performance in terms of prediction accuracy in bankruptcy prediction and misclassification cost.

On the performance of learning machines for bankruptcy detection

Predicting the financial health of companies is a problem of great importance to various stakeholders in the increasingly globalized economy. We apply several learning machines methods to the problem of bankruptcy prediction of private companies. Financial data obtained from Diana, a database containing 780,000 financial statements of French companies, are used to perform experiments. Classification accuracy is evaluated with respect to artificial neural networks, linear genetic programming and support vector machines. We analyze both type I (bankrupted companies misclassified as healthy) and type II (healthy companies misclassified as bankrupted) errors on three datasets containing balanced and unbalanced class distribution. Linear genetic programming has the best accuracy in the balanced data while support vector machines is more stable for the unbalanced dataset. Our results, though preliminary in nature, demonstrate the tremendous potential of using learning machines in solving important economics problems such as predicting bankruptcy with accuracy