Journal of Management and Business Solutions 4 Serial Number 17 2026 01 01 The Role of Media in Identifying and Predicting Financial Fraud in Companies Listed on the Tehran Stock Exchange The Role of Media in Identifying and Predicting Financial Fraud in Companies Listed on the Tehran Stock Exchange 1 26 EN Journal Article 2025 09 03 <p>The aim of this study was to evaluate the role of media in identifying and predicting financial fraud in companies listed on the Tehran Stock Exchange using machine learning algorithms. In the first stage, text-mining analysis of official news websites, together with content screening of studies relevant to the research objective, was employed to identify effective media-related criteria for predicting the likelihood of financial fraud. After assessing the validity and reliability of the identified dimensions, machine learning algorithms were used to evaluate the effectiveness of these criteria in identifying and predicting corporate financial fraud. For this purpose, data on companies listed on the Tehran Stock Exchange covering the period from 2019 to 2025 were collected. Following initial preprocessing, data from 95 companies were classified, based on the Modified Beneish Model, into non-fraudulent samples and samples with a likelihood of financial fraud. Given the class imbalance and considerable overlap between the non-fraudulent observations and those indicating a likelihood of financial fraud, the Synthetic Minority Over-sampling Technique (SMOTE) was applied exclusively to the training dataset to prevent data leakage. Subsequently, to ensure a fair assessment and evaluate the robustness of machine learning models encompassing both supervised and unsupervised approaches, and with the objective of identifying an effective preliminary algorithm for detecting signals derived from analytical media in predicting the likelihood of fraud among companies listed on the Tehran Stock Exchange, a systematic approach based on grid search and repeated five-fold cross-validation with five repetitions (R5FCV) was employed. The results indicated that, among the various supervised and unsupervised learning models examined, supervised learning models—particularly logistic regression—achieved superior performance metrics and demonstrated greater accuracy and discriminatory power in predicting the likelihood of financial fraud based on operational and process-related measures derived from media functions among companies listed on the Tehran Stock Exchange. These findings confirm that the extracted measures possess meaningful analytical capacity for predicting financial fraud and may serve as complementary tools for capital market analysts and regulatory authorities.</p> https://journalmbs.com/index.php/jmbs/article/download/421/409