This study aims to detect and analyze potential fraud in inventory management using an anomaly detection approach based on event log data at CV Putera Bumi’s warehouse system. The main issue faced by the company is the discrepancy between system data and physical stock conditions due to possible record manipulation and weak internal control. The research employs a combination of process mining using the Alpha Algorithm to reconstruct actual process flows and the Isolation Forest method to detect anomalies without requiring explicit fraud labels. The dataset consists of warehouse event logs recorded over the past 12 months. The findings reveal that the hybrid approach effectively identifies deviations from standard operating procedures and detects potential fraudulent patterns such as after-hours transactions, repeated correction activities, and process skipping. Compared to conventional audit or rule-based methods, this data-driven approach proves more adaptive in recognizing contextual anomalies automatically. The study provides practical contributions for strengthening internal control systems and theoretical insights into the application of unsupervised learning for fraud detection in event log data.