HEFZ – RNNLSTM: An Ingenious Deep Learning Hybrid Model for Ensemble-Based Prediction of Potential Fishing Zone Areas in the Indian Ocean

Abstract

Many millions of people call the Indian Gujarat subcontinent home, and it's famous for its long coastline (more than 1,400 kilometers long). Fishing is crucial to the survival of their ecosystem. Finding fish can be time-consuming and resource-intensive, which can drive up operational costs and reduce profits. Artificial intelligence (AI) algorithms have allowed for significant advancements in the design of advanced algorithms for the forecast of good fishing spots, both in terms of superior accuracy (Acy) and reduced time. Still, it's never an easy task to foretell where PFZs might appear. According to mitigate these issues, this study introduced a model for innovative hybrid PFZ prediction by combining remote sensing datasets with conventional dataset methodology. The proposed Model combines deep convolutional layers with recurrent neural networks based on the flitter adam optimization of long short-term memory (FA-LSTM) (RNN). Chlorophyll, sea surface temperature (SST), GPS location of last fishing location, fish living temperature, and other features are all removed using convolutional layers before FB-LTSM is used to predict future fishing spots. Extensive experiments are conducted with satellite data from NASA's Ocean colour web, and TensorFlow 1.18 with Keras API is used to implement the necessary infrastructure. The performance metrics are compared with other existing intelligent learning models, including F1-score, sensitivity, sensitivity, recall, specificity, and precision (Pscn), as well as accuracy (Acy),. Our data shows that the proposed Model (94 percent prediction Acy) outperforms the state-of-the-art algorithms, making it an ideal candidate for use in developing a smart system for improved PFZ prediction.