TGFish: A Temporal Graph Convolutional Network for Fishing Effort Distribution Prediction

Yufei Jiang,Yuqi Shi,Siyuan Lin,Feng Hong
DOI: https://doi.org/10.1109/oceans51537.2024.10682238
2024-01-01
Abstract:Predicting the distribution of fishing effort is essential for effective fisheries management, allowing for the creation of fishing plans and the preservation of marine biological resources. However, this prediction faces three main challenges: the unquantified impact of hydrological factors on fishing effort distribution, as well as the spatial and temporal patterns of fishing vessel behavior. In this study, we aim to measure the correlation between hydrological factor fields and fishing effort distribution. From this analysis, we propose a system called TGFish, which can predict day-to-day fishing effort distribution for the upcoming week. TGFish integrates graph convolution neural networks to extract features from fishing effort distribution and hydrological factor fields, addressing spatial constraints, and LSTM to handle temporal constraints. We use the VMS dataset from 1,899 trawling fishing vessels in the East China Sea between September 2015 and May 2017. The dataset from September 2015 to May 2016 is used for correlation analysis and model training, while the dataset from September 2016 to May 2017 is employed to evaluate prediction accuracy. Extensive experimentation demonstrates that TGFish effectively predicts the day-level distribution of fishing effort for the upcoming week. The results confirm the effectiveness and reliability of our proposed system.
What problem does this paper attempt to address?