Artificial Intelligence in Corporate Sustainability: Using LSTM and GRU for Going Concern Prediction

Abstract

“Going concern” is a professional term in the domain of accounting and auditing. The issuance of appropriate audit opinions by certified public accountants (CPAs) and auditors is critical to companies as a going concern, as misjudgment and/or failure to identify the probability of bankruptcy can cause heavy losses to stakeholders and affect corporate sustainability. In the era of artificial intelligence (AI), deep learning algorithms are widely used by practitioners, and academic research is also gradually embarking on projects in various domains. However, the use of deep learning algorithms in the prediction of going concern remains limited. In contrast to those in the literature, this study uses long short-term memory (LSTM) and gated recurrent unit (GRU) for learning and training, in order to construct effective and highly accurate going-concern prediction models. The sample pool consists of the Taiwan Stock Exchange Corporation (TWSE) and the Taipei Exchange (TPEx) listed companies in 2004–2019, including 86 companies with going concern doubt and 172 companies without going concern doubt. In other words, 258 companies in total are sampled. There are 20 research variables, comprising 16 financial variables and 4 non-financial variables. The results are based on performance indicators such as accuracy, precision, recall/sensitivity, specificity, F1-scores, and Type I and Type II error rates, and both the LSTM and GRU models perform well. As far as accuracy is concerned, the LSTM model reports 96.15% accuracy while GRU shows 94.23% accuracy.

Using Hybrid Artificial Intelligence and Machine Learning Technologies for Sustainability in Going-Concern Prediction

“…Jan [16] uses the classification and regression tree (CART), deep neural network (DNN), and recurrent neural network (RNN) to construct going-concern prediction models, among which, the prediction accuracy of the CART-RNN model is the highest (95.28%). Chi and Chu [23] use long short-term memory (LSTM) and gated recurrent unit (GRU) to construct going-concern prediction models, among which, the prediction accuracy of the LSTM model is the highest (96.15%). In these important studies of going-concern decisions by machine learning or deep learning, the accuracy of the constructed going-concern prediction models is higher than 80%, which greatly inspires this study.…”

“…The confusion matrix [16,23] is also used in this study, the indicators of confusion matrix are accuracy, precision as Equation ( 14), recall (sensitivity) as Equation (15), and F1-score as Equation ( 16). precision = ture postive/(ture postive + f alse postive) (14) recall = ture positive/(ture postive + f alse negative) The sample period in this study is 20 years from 2000 to 2019, listed and OTC (overthe-counter) companies issued with audit opinions of going-concern doubts are taken as the sample subjects, and the companies issued with audit opinions of going-concern doubts in the first year of the study period are selected as the samples.…”

“…Natural log of total assets Ln (total assets) Chen [1]; Chen and Lee [5]; Goo et al [6]; Jan [33]; Chen and Jhuang [34] X2 Debt ratio Total liabilities/total assets Chen [1]; Chen and Lee [5]; Goo et al [6]; Jan [16]; Chen and Shen [22]; Jan [33]; Chen and Jhuang [34]; Jan [35]; Jan [36] X3 Quick ratio Quick assets/current liabilities Chen [1]; Jan [16]; Pawełek [21]; Chen and Shen [22]; Jan [33]; Chen and Jhuang [34]; Jan [35]; Jan [36] X4 Current ratio Current assets/current liabilities Chen [1]; Chen and Lee [5]; Goo et al [6]; Jan [16]; Chen and Shen [22]; Chi and Chu [23]; Jan [33]; Jan [35]; Jan [36] X5 Total liabilities/ stockholders' equity Total liabilities/stockholders' equity Chen [1]; Jan [16]; Chen and Shen [22]; Jan [36] X6 Current liabilities/total liabilities Current liabilities/total liabilities Chen and Lee [5]; Jan [16]; Chi and Chu [23]; Jan [35] X7 Current assets/total assets Current assets/total assets Chen [1]; Chen and Lee [5]; Chi and Chu [23]; Jan …”

A Stochastic Multivariate Irregularly Sampled Time Series Imputation Method for Electronic Health Records

“…In the original version, the model in [10] uses an integrated GRU cell to incorporate the Bayesian update with observed values. However, in this study, we used an LSTM cell to perform the Bayesian update since it provides a more accurate estimation [25]. However, the missing values are not updated since they do not have any effect on the hidden layers.…”

Prediction of outpatients with conjunctivitis in Xinjiang based on LSTM and GRU models

“…GRU networks are proposed on the basis of LSTM networks, which also take into account long-term dependencies. Compared to the LSTM, the GRU has one less gate function and requires fewer parameters, therefore a shorter training time is required [ 23 ]. There are similarities and differences between LSTM and GRU methods, but it is impossible to judge theoretically which method is better.…”