Fishing Spot Prediction by Sea Temperature Pattern Learning

Determination of appropriate fishing spots is one of the most important activities in the fishing industry. Inspired by the approach followed by fishermen to determine fishing spots, this paper presents a new machine-learning method for uncovering oceanographic patterns related to good fishing spots. Our method uses a sea temperature map as the input, extracts sea temperature patterns from arbitrary points on the map, and evaluates whether the patterns correspond to good fishing spots by using two machine learning techniques; one-class support vector machine (SVM) and spectral clustering. We evaluated the efficiency of our method using fishery data on neon flying squid.