Evaluating Machine Learning Algorithms for Financial Fraud Detection: Insights from Indonesia

The study utilized Multiple Linear Regression along with advanced classification algorithms such as Logistic Regression, K-Nearest Neighbors (KNN), Support Vector Machine (SVM), Decision Tree, and Random Forest, to detect financial statement fraud. Model performance was evaluated using key metrics, including precision, recall, accuracy, and F1-Score. The analysis also identified significant indicators of fraud, such as Accounts Receivable Turnover, Days Outstanding Accounts Receivable, Days Payables Outstanding, Logarithm of Gross Profit, Gross Profit Margin, Inventory to Sales Ratio, and Total Asset Turnover. Among the models, Random Forest emerged as the most effective algorithm, consistently outperforming others on both training and testing datasets. Logistic Regression and SVM demonstrated strong reliability, whereas KNN and Decision Tree faced overfitting challenges, limiting their practical application. These findings emphasize the critical need for enhanced fraud detection frameworks, leveraging machine learning algorithms like Random Forest to identify fraud patterns effectively. The study highlights the importance of strengthening internal controls, implementing targeted fraud detection measures, and promoting regulatory improvements to enhance transparency and financial accountability.