K-Means Clustering for Information Dissemination of Fishing Surveillance

Abstract

The Portuguese Navy is responsible for monitoring the largest Exclusive Economic Zone in Europe. The most captured species in this area are Scomber colias and Trachurus trachurus, commonly called Mackerel and Horse Mackerel, respectively. One of the Navy’s missions is pursuing actions of fishing surveillance to verify the compliance of proceedings with the species’ fishing activity regulation. This monitoring actions originate data that represents a sample of the fishing activity in the area. The collected data, analysed with adequate data mining techniques, makes it possible to extract useful information to better understand the fishing activity related to Mackerel and Horse Mackerel, even if the full data set cannot be disclosed. With this in mind the authors used a non-supervised learning technique, the K-Means algorithm, which grouped data in clusters by its similarity and made a summarized description of each cluster with the purpose of releasing a general overview of such records. The information obtained from the clusters led the authors to deepen the study by performing a comparison of the monthly average quantity recorded per vessel for the two species in order to infer about the relation between captured quantity Mackerel and Horse Mackerel over time.