Loading...

Automatic Financial Statement Defect Detection of Publicly-traded Companies

Ashoori, Maedeh | 2024

0 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 57784 (01)
  4. University: Sharif University of Technology
  5. Department: Industrial Engineering
  6. Advisor(s): Habibi, Moslem
  7. Abstract:
  8. Detecting defects in financial statements is crucial because such defects can impact the transparency, accuracy, and reliability of financial reports and influence the decision-making of stakeholders. This study aims to provide a novel approach for identifying defects in the financial statements of companies listed on the Tehran Stock Exchange. In this research, a defect is defined as a difference in data items between audited and unaudited financial statements, which may indicate either unintentional or intentional deviations or signal a potential fraud. The data for this study was automatically collected from the Codal system and includes both audited and unaudited financial reports in the form of balance sheets and income statements. The data preprocessing process involved calculating the percentage of discrepancies in financial variables between the audited and unaudited financial statements, along with other measures to enhance the quality of the input data for the models. Additionally, the collected data was categorized and labeled based on industry to identify defects in the reports. In the modeling phase, the Random Forest algorithm was used as the primary tool for detecting defects. This model was chosen due to its ability to handle large datasets, its high interpretability, and its capacity to reduce overfitting. The algorithm was optimized by using financial ratios and analyzing trends in the changes of financial variables over different periods to improve the accuracy of deficiency detection. The results of this research showed that the optimized model, with a sensitivity measure of 86%, was able to identify defects and patterns related to identifying defects and suspicious reports. Throughout the research process, various methods and approaches were utilized. Specifically, financial ratios were added as additional features to the model, and through careful analysis and optimization of the model's parameters, we were able to improve the model’s performance. Moreover, analyzing the trends in changes of financial variables over specified time periods led to significant improvements in the model’s performance. These comprehensive and multi-faceted approaches were highly influential in achieving the best results for identifying suspicious reports with higher accuracy
  9. Keywords:
  10. Financial Fraud ; Financial Fraud Detection (FFD) ; Machine Learning ; Tehran Stock Exchange ; Financial Statements ; Defect Detection

 Digital Object List

 Bookmark

No TOC