Marine Rain Prediction using Enhanced Deep Learning techniques

Abstract

Marine weather forecasting is a critical aspect of maritime operations, impacting safety,efficiency and planning. This research introduces three enhanced techniques: Ridge Regularized Imputed Scaled Clipping Normalization (RRISCN), Fisher Kernel Target Projective (FKTP), and Reinforced LSTM with Partial Differential Quantum Hamiltonian Neural Network (ReLSTM-PDQHNN). These techniques are applied to various sample data instances from a dataset, focusing on pre-processing, feature selection, and data classification to predict rain or newlineno rain conditions. The objective of this research is to evaluate the performance of these techniques in marine weather forecasting, specifically in predicting rain or no rain conditions. newlineThe research aims to assess the effectiveness of each method in handling various challenges newlinesuch as multi-collinearity, missing data, and non-linear relationships, and to identify the most suitable approach for accurate and reliable weather predictions. newlineThe RRISCN method employs ridge regularization to address multi-collinearity and imputation techniques to handle missing data. It also incorporates clipping normalization to scale the data, ensuring robustness against outliers. The FKTP technique utilizes Fisher Kernel newlinetarget projection to capture the underlying structure of the data, enhancing feature representation and selection. The ReLSTM-PDQHNN approach combines the power of Reinforced LSTM with Partial Differential Quantum Hamiltonian Neural Network to model complex temporal dependencies and non-linear relationships in the data. The research evaluates the performance of these techniques, using metrics such as accuracy, error rate and prediction time. Experimental newlineresults demonstrate the effectiveness of each method in marine weather forecasting, with the ReLSTM-PDQHNN technique outperforming the others in terms of predictive accuracy and newlinerobustness in varying weather conditions.