Predicting Going Concern of Companies Using Text Mining and Data Mining Approaches

Document Type : Original Article

Authors

• Hamid Abbaskhani ¹
• Asgar Pakmaram ¹
• nader rezaei ²
• Jamal Bahri Sales ³

¹ Department of Accounting, Bonab Branch, Islamic Azad University, Bonab, Iran

² Department of Accounting, Bonab Branch, Islamic Azad University, Bonab, Iran.

³ Department of Accounting, Urmia Branch, Islamic Azad University, Urmia, Iran

10.22067/ijaaf.2022.78600.1217

Abstract

The linguistic features of the information provided by the business unit can facilitate the achievement of the objectives of conveying economic facts. Thus, in recent years, such features have always been taken into account in accounting and behavioral finance studies. Therefore, the purpose of this study is to determine the ability to predict the going concern of companies using structured and unstructured data, as well as any changes that occur in it because of adding unstructured variables to purely data mining models. In addition, if the results are different, are the difference significant or non-significant? The study period was from 2012 to 2021 and the study sample included 54 companies listed on Tehran Stock Exchange. The tone of the auditor's report was measured using the Mayew et al. (2015) and the Visvanathan (2021) models. The MAXQDA 20 text analysis software and the Loughran and MacDonald (2015) dictionary were also used to process the data. Data analysis and hypothesis testing were done using the logit regression model and the Vuong test. The results of the test of the first hypothesis indicate that the text-based method model has a higher coefficient of determination than the data-based method model, and the test of the second hypothesis shows that there is a significant difference in the exponential explanatory power of the data-based method model and the data-based method model in companies.

Keywords

• Auditor Report
• Sentiment Analysis
• Going Concern
• Text Mining

Main Subjects

• Accounting

---