A Heuristic Neural Network Approach for Underwater Parametric Prediction at Bay of Bengal

Abstract

The fundamental ocean properties are vital in predicting the earth’s climate. Such predictions are essential for human life in various terrestrial applications. Therefore, to forecast the parametric variations in the Bay of Bengal with less prediction error, a Heuristic Neural Network (HNN) is proposed by optimizing the solutions of Long Short-Term Memory (LSTM) using swarm intelligence by Repeated Iterative Technique (RIT). The existing algorithms have proved better accuracy only during the sea surface prediction of temperature. The proposed HNN model predicts dominant parameters of the Bay of Bengal for its horizontal and vertical variations from 10 to 2000 m depth. The efficiency of the proposed HNN model is verified by comparing the results with recent prediction algorithms in fusion with LSTM, from which the outcome reveals that the proposed HNN model obtains 98.9% for temperature prediction, 99% for pressure, 98.8% for salinity, and 98.1% for density prediction. The results prove overall good accuracy of forecast compared to the existing prediction techniques.