Document Type : Research Paper
Abstract
Purpose: The aim of this study is to predict the behavior of auditors when confronted with fraud and financial misconduct and to identify the factors influencing their fraudulent behavior.
Method: This research adopts a quantitative approach, with the statistical population consisting of auditors from auditing firms in Tehran. The sample size was determined to 256 individuals using Morgan's table. Data were collected through a questionnaire designed to identify factors affecting auditors' fraudulent behavior. For designing the neural network model, Python software was utilized. Key steps included data processing, division into training, testing, and validation datasets, selection of the network type, and weight adjustment using backpropagation algorithms. Data processing involved detrending and normalization, and the data were divided into training, testing, and validation sets.
Findings: The analysis results indicate a high accuracy of the neural network model in identifying auditors with and without a history of fraud, achieving an overall accuracy of 96.15%. This high accuracy suggests that the proposed model can effectively identify the factors influencing auditors' fraudulent behavior and provide precise predictions.
Conclusion: This study demonstrates that the use of neural network models and analytical technologies can significantly assist auditing firms in strengthening their supervisory systems. By implementing these models, the possibility of preventing financial misconduct increases, creating a more transparent and secure working environment.
Knowledge Contribution: The knowledge contribution of this study lies in its use of neural networks to predict auditors' behavior in response to fraud and financial misconduct, enhancing the accuracy and speed of supervisory decision-making.
Main Subjects
)