THRESHOLD-BASED ANOMALY DETECTION IN DRY BULK CARGO VOLUME USING SIMULATED LSTM AUTOENCODER RECONSTRUCTION ERROR

Abstract

This study addresses the challenge of detecting anomalies in annual dry bulk cargo volumes at a major Indonesian port by simulating the reconstruction error typically produced by an LSTM Autoencoder model. Instead of applying deep learning directly, the research utilizes a statistical approximation involving a three-year centered moving average to emulate the expected cargo pattern. The absolute deviation between actual and smoothed values is treated as simulated reconstruction error. A statistical threshold is then calculated based on the mean and standard deviation of these errors to distinguish normal years from anomalous ones. Results indicate that only the year 2023 exceeded the anomaly threshold, suggesting significant irregularity in cargo flow during that period. The proposed method offers a practical and interpretable framework for anomaly detection, particularly in data environments lacking access to machine learning infrastructure. This approach enables port operators and planners to monitor unusual volume fluctuations efficiently and provides a foundation for further integration of data-driven risk management systems. Keywords: Dry Bulk Cargo, Anomaly Detection, Reconstruction Error, LSTM Autoencoder, Moving Average, Threshold Classification, Port Operations.