Data Mining approach to identify PFZ(Potential Fishing Zone) using Oceanographic Data

Abstract

Oceanographic factors (biological, physical and chemical) and fisherman’s expertise are the two basic ways for determining potential fishing zones. There are many disadvantages with this approach when it comes to determining the exact potential fishing zones(PFZs) spatially and temporally. In our proposed framework, we have come up with a data mining approach for identifying PFZ in Indian ocean. We procured spatio -temporal images from Copernicus Online Data Access (CODA) and then extracted characteristics, Sea Surface Temperature (SST) and Sea Surface Chlorophyll(SSC) with the help of Sentinel Application Platform(SNAP). The results of this extraction method were used as Training data in the classification process, which was then used to PFZs. During the classification process, we utilized ensemble learning approach consists of different algorithms such as KNN, Naïve-Bayes, KNN classifier-5, KNN-classifier-10, Support Vector Machine (SVM), logistic regression, decision tree. The result gave an average accuracy of 80%, which showed that the proposed framework can be used effectively to determine PFZs. To validate the framework, we followed the process of cross-validation with the labelled data. The results showed that the proposed data mining framework predicted the correct values.