<![CDATA[In the digital era, the rapid growth of electronic payment transactions using debit cards has been accompanied by an increasing risk of anomalies and fraudulent activities. Identifying suspicious transactions has become crucial to ensure system security and maintain user trust. The high volume of transactions processed through switching systems in Indonesia poses significant challenges for operational teams in detecting anomalous patterns effectively. This study aims to identify anomalous debit payment transactions within switching networks by comparing three classification methods, namely Naive Bayes, K-Nearest Neighbors (K-NN), and Decision Tree. The dataset used consists of sampled daily transaction data obtained from operational monitoring, which is analyzed based on predefined transaction matrices developed by operational teams as indicators of anomaly detection. The evaluation of model performance is conducted using key metrics, including accuracy, precision, and recall, to determine the most effective classification method. The results show that machine learning-based classification significantly improves the accuracy and efficiency of anomaly detection compared to manual analysis. Furthermore, the integration of data mining techniques with operational transaction matrices provides a structured and practical approach for early anomaly identification. This approach not only enhances the effectiveness of transaction monitoring but also strengthens fraud prevention mechanisms and supports more informed and data-driven decision-making processes within switching companies.]]>
