A comparative analysis of artificial neural networks topologies in default forecasting
AbstractArtificial neural networks are widely used in many classification tasks. One of them can be the classification of potential debtors into healthy and unsound due to their ability to pay back the debt in contractual time due to the ex-post financial information (financial ratios). Though the entry data may vary, depending on the applied approach, data limitations or the choice of ratios selection, it is the topology of the neural networks that has the biggest input on the outcome. The objective of the paper is to investigate the use of different artificial neural networks (NN) structures in the process of classification of banks’ potential clients. Topologies and architectures of neural network are presented through literature review and an experimental case study implementing compares the results of a supervised learning multi-layer perceptron feed-forward neural network (MLP-FF) with the one based on back-propagation (denoted as MLP-BP). The paper uses a cross-national sample of 300 companies applying for credit to an international bank operating in Poland. Results of those different methods are juxtaposed, and their performance compared to present which architectures are more efficient than others in debtors’ classification process.
|Publication size in sheets||0.5|
|Book||Szkutnik Włodzimierz, Sączewska-Piotrowska Anna, Hadaś-Dyduch Monika, Acedański Jan (eds.): 10th International Scientific Conference: Analysis of International Relations 2018. Methods and Models of Regional Development. Summer Edition. Conference Proceedings, 2018, Wydawnictwo Uniwersytetu Ekonomicznego w Katowicach, ISBN 978-83-7875-456-5, [978-83-7875-455-8], 145 p.|
|Keywords in English||credit risk, default, neural networks|
|Score||= 20.0, 24-06-2020, ChapterFromConference|
|Publication indicators||= 0|
* presented citation count is obtained through Internet information analysis and it is close to the number calculated by the Publish or Perish system.