Using machine learning to detect misstatements

Deep Learning Based on Hierarchical Self‐Attention for Finance Distress Prediction Incorporating Text

“…The quantitative financial indicators are downloaded from the China Stock Market and Accounting Study database (CSMAR). Based on previous researches [ 5 , 10 , 12 , 24 ], 48 financial indicators are taken into account, including solvency, ratio structure, operation, profitability, cash flow, risk, development, and the index of per share. Solvency and cash flow describe a company's ability to repay short-term and long-term debts to prevent bankruptcy.…”

“…Most previous researches focused on the application of machine learning methods to gain insights into financial indicators as clues to detect financial risk. For model construction, on one hand, classic statistical and machine learning methods are applied in feature engineering and classification, such as Naïve Bayesian [ 5 , 6 ], Support Vector Machine (SVM) [ 2 , 7 , 8 ], and ensemble learning including decision trees based Gradient Boosting Decision Tree (GBDT) [ 9 – 12 ], Random Forest (RF) [ 13 , 14 ], eXtreme Gradient Boosting (XGB) [ 13 , 15 ], and Adaptive Boosting (AdaBoost) [ 16 , 17 ]. On the other hand, various deep learning models are also employed for modeling [ 18 ], such as Genetic Algorithm (GA) [ 6 , 19 ], Convolutional Neural Network (CNN) [ 20 , 21 ], and Self Organizing Map (SOM) [ 22 ].…”

“…With the development of artificial intelligence (AI), experts in the field of finance and accounting devote themselves to integrating heterogeneous massive amounts of information by the devices with powerful computing capabilities to predict financial distress more accurately [ 12 , 14 , 17 ]. Relevant research proved that text fusion benefits more accurate identification of financial distress [ 4 , 5 , 25 ].…”

Using Machine Learning to Measure Conservatism

“…the CV training, later years assigned to validation, and final years in the sample assigned to the test sample (Bao et al 2020;Bertomeu et al 2020). This research design avoids using information to build a prediction that may not be known yet by the time the prediction is made.…”

“…Studies such as Perols (2011) and Perols et al (2017) are among the first to introduce machine learning methods in accounting. Recent studies have also examined the use of machine learning for textual analysis (Li 2010;Frankel et al 2016), to analyze audit quality (Yang et al 2020;Hu et al 2020), to compute more accurate accruals (Ding et al 2020), to examine the evolution of the relevance of accounting (Barth et al 2019) or to predict SEC enforcement actions (Bao et al 2020) and misstatements (Bertomeu et al 2020). 3 This literature focuses on prediction models.…”