Fusion Improved DPC and LOF Algorithms for Anomaly Detection in Buoy Position Data

Buoy position data is a key source of information for studying buoy drift characteristics and detecting displacement. However, when collecting data using a buoy telemetry system, abnormal position data often arises due to signal transmission issues and external natural factors, which can affect the effective use of buoy data. To address this issue, a buoy position data Anomaly detection algorithm, SDPC-LOF, is proposed, which fusion the improved density peak clustering algorithm based on shared nearest neighbor(SDPC) with local outlier factor(LOF) algorithm. SDPC optimizes the local density definition and sample assignment mechanism in DPC by introducing SNN and a multi-cluster fusion strategy, thereby improving clustering performance and conducting detailed classification of complex buoy data. Subsequently, LOF is employed to detect outliers in the buoy position data and remove local outliers from the dataset, effectively processing anomalous buoy position data. The practical application in Port of Xiamen shows that SDPC-LOF algorithm can detect and deal with the abnormal points of the buoy position data effectively.