Investigating the quality and sustainability of accruals: text mining approach

This study investigates the quality and sustainability of accruals of listed companies in Tehran Stock Exchange through qualitative and unstructured reports in the period 2011-2017. Statistical technique used to test hypotheses, Multivariate regression of panel data type using The STATA software was used to measure accruals in board reports (accrual accruals) using support vector regression technique in Python software. The accruals in board reports (textual accruals) were measured using support vector regression technique in Python software. The research findings show that textual accruals are more explanatory power than actual accruals (company level accruals) for measuring accrual quality. Also textual accruals estimated by the support vector regression technique are more stable than non-textual accruals that are not determined by this method using text mining techniques (especially support vector regression) is able to transform qualitative and unstructured disclosures such as Board of Directors Reports to an estimate of the fundamental elements (quality and sustainability of accruals) at the company level.