Home // ICIW 2018, The Thirteenth International Conference on Internet and Web Applications and Services // View article

Improving Default Risk Information System with TensorFlow

Authors:
Claudio Augusto Silveira Lelis
Andre Luiz Silveira Lopardi

Keywords: Indicators; Financial Management; Knowledge-based Decision-making; Default Prediction Model; TensorFlow.

Abstract:
The decision process is essential in the granting of credit. The right decision can be critical in reducing financial losses. DeRis (Default Risk Information System) is an information system designed to support activities in the management of default risk. The main component is a predictive model of default based on indicators. Currently, the system has been improved allowing models of the TensorFlow tool. Based on real datasets, the default model and the predictive models of the TensorFlow tool was evaluated for different types of indicators. The results show that a model optimization was possible through the adjustment of the hyperparameters offered by TensorFlow, with 240 distinct combinations being tested between these hyperparameters. Although the results are associated with the data and the design of the experiment conducted, they were considered positive and promising for future work.

Pages: 24 to 29

Copyright: Copyright (c) IARIA, 2018

Publication date: July 22, 2018

Published in: conference

ISSN: 2308-3972

ISBN: 978-1-61208-651-4

Location: Barcelona, Spain

Dates: from July 22, 2018 to July 26, 2018