Ocean Data Anomaly Detection Algorithm Based on Improved k-medoids

Abstract

The topic of abnormal data mining in ocean Argo buoy monitoring data is studied. Firstly, dense regions were established in K-MEDOIDS clustering algorithm with the help of density accessibility of density clustering. Based on dynamic layer number, a new calculation method of domain radius and density was proposed, and the initial clustering center was selected with both considering density and similarity; At the same time, an anomaly detection algorithm is proposed, which the criterion to judge marine anomaly data is based on the result of clustering combined with point sets in dense regions. Experimental verification was carried out on the actual and artificial simulated data sets, the results show that the clustering performance and anomaly detection are improved compared with the comparison algorithm.