Presenting a Fraud Prediction Model Based on Artificial Intelligence (Naïve Bayes)

After the technological revolution and the advent of the Internet, the rise of artificial intelligence has had a particularly profound impact on various fields, including accounting. In fact, artificial intelligence-based models are able to perform tasks such as detecting or predicting fraud by analyzing data, which significantly reduces human error and provides more reliable outcomes. In this context, the objective of the present study is to examine the efficiency and effectiveness of the "Naive Bayes" method in predicting fraud.

The present study is applied in terms of its purpose and causal in terms of its nature and methodology. It is also retrospective regarding the characteristics and direction of the data, utilizing historical information. This study aims to predict fraud in the financial statements of companies listed on the Tehran Stock Exchange. To select a statistical sample, only companies engaged in production activities were included, which means that financial institutions and banks were excluded from the sample. On the other hand, the selected companies must have a fiscal year that ends in March. It is important to note that the research period spans from 2016 to 2022. Necessary measures were implemented to gather theoretical resources through library research, as well as to collect the data needed to test the hypotheses using archival methods and by consulting the Stock Exchange website. In this research, information related to the time series of the total index was collected from the official website of the Stock Exchange Company. Excel software was utilized to organize the data and perform basic calculations on the raw data, while Python was employed to analyze the data and develop artificial intelligence models.

The method studied in this study demonstrates a strong capacity to predict fraud based on the variables present in the financial statements of publicly listed companies, achieving an accuracy rate of 84%, a precision rate of 84%, and a recall rate of 98%.

Fixed assets and capital variables significantly influence the occurrence of fraud. It is anticipated that any changes in the company's income and profitability levels will also result in corresponding adjustments in the volume of fixed assets and available capital. Therefore, companies can more easily commit fraud to manipulate their performance, given the accessibility and simplicity of altering the reporting of these variables. The results obtained are consistent with the definitions provided by the Association of Official Auditors, which states that changes in assets are a significant variable contributing to the occurrence of fraud in various companies. Therefore, fluctuations in asset and capital levels can serve as significant indicators of potential fraud.