INSOLVENCY PREDICTION OF AUSTRALIAN CONSTRUCTION COMPANIES USING DEEP LEARNING WITH BIDIRECTIONAL LSTM AUTOENCODER

Abstract

Business insolvency in the building and construction industry is a major concern on a worldwide scale, and it is particularly pervasive in the Australian construction industry. Many Australian construction companies frequently uses high levels of borrowing and poor profit margins, which increases the likelihood of insolvency. This paper develops a novel, intelligent insolvency prediction model for the Australian construction companies. The proposed framework with bidirectional long short-term memory (BiLSTM) models and autoencoder techniques contains not only the financial variables but also other important indicators that are linked to the features of the sector that have previously been disregarded. Finally, numerical experiments show that the proposed neural network model outperforms several existing models for predicting the insolvency of construction companies.