University of Qom Journal of Data Analytics and Intelligent Decision-making 3115-8161 2 3 2026 07 01 From Prediction to Explanation: Earnings Management Analysis Using Explainable Machine Learning and the SHAP Approach 82 121 4826 10.22091/jdaid.2026.16254.1070 EN Saeed Alipour Department of Accounting, Ard.C., Islamic Azad University, Ardabil, Iran 0000-0002-1802-4460 Mehdi Fazli Department of Mathematics, Ard.C., Islamic Azad University, Ardabil, Iran 0000-0002-4089-1827 Journal Article 2026 06 05 Objective: This study aims to develop and evaluate an explainable machine learning framework for detecting earnings management and to compare its predictive performance with traditional accrual-based approaches in the Iranian capital market. Beyond improving prediction accuracy, the study explains the nonlinear relationships between financial, corporate governance, and market variables associated with earnings management using explainable artificial intelligence techniques.<br /><br />Method: This applied study adopts a descriptive-analytical design based on machine learning and explainable artificial intelligence. The statistical population consists of firms listed on the Tehran Stock Exchange during 2016–2025. After applying the sample selection criteria, 1,620 firm-year observations were analyzed. Five machine learning algorithms—XGBoost, Random Forest, Support Vector Machine, Decision Tree, and Logistic Regression—were evaluated and compared with a traditional accrual-based approach. Model interpretation was performed using SHAP (SHapley Additive exPlanations), while predictive performance was assessed using Accuracy, Precision, Recall, F1-score, and AUC.<br /><br />Results: XGBoost achieved the highest predictive performance, with an AUC of 0.90 and an F1-score of 0.86, significantly outperforming the traditional accrual-based approach and the other machine learning models. SHAP analysis revealed that operating cash flow volatility, financial leverage, market-to-book ratio, audit quality, and other financial and governance variables exert substantial nonlinear effects on earnings management prediction.<br /><br />Conclusions: The findings demonstrate that explainable machine learning not only improves predictive performance but also enhances model transparency and theoretical interpretation. The proposed framework provides valuable analytical support for auditors, regulators, investors, and corporate decision-makers seeking to strengthen financial reporting quality, regulatory oversight, and earnings management monitoring. https://jdaid.qom.ac.ir/article_4826_912ea27d67e564630be31505830a97dd.pdf