Development of an Information System with a Machine Learning Approach for Anomaly Detection in Hotel Transactions
Abstrak
Hotel customer transaction data contains valuable information for understanding consumer behavior, but anomalies such as unusual guest numbers or booking patterns can disrupt analysis accuracy. This research develops a machine learning-based anomaly detection feature integrated into the hotel information system to identify unusual transactions and provide insights through customer segmentation. The system operates automatically when transactions are recorded, performing two stages: anomaly detection and segmentation classification. Anomaly detection uses the Isolation Forest algorithm, which successfully separates normal and anomalous data (visual gap ~48). Detected anomalies are classified using K-Means and Random Forest. K-Means shows overlapping clustering (Silhouette Score 0.3372; DBI 1.0305), while Random Forest achieves 96% accuracy, proving its effectiveness in classification. System testing through Black Box Testing, Stress Testing, and User Acceptance Testing shows the system functions well, handles high loads, and receives positive feedback from users. The system not only records transactions but also provides real-time analysis of unusual customer behaviors, supporting more accurate decision-making.
