Nearshore significant wave height prediction based on MIC-LSTM model

The near-shore sea area, which is a region of frequent human activities, is important to explore the causes of meteorological factors that are close to human activities in it. A prediction model based on the fusion of maximum information coefficient (MIC) and long and short-term memory network (LSTM) is proposed to predict the significant wave heights of nearshore ocean hydrographic data at Zhapo Harbor, Hailingshan Island, Yangjiang City, Guangdong Province, China, and the public data of deep-sea buoy 44,013 of the National Data Buoy Center (NDBC) in the U.S. The model is applied to the prediction of significant wave heights by combining with support vector regression (SVR), LSTM combined with Residual Network (ResNet), and XGBoost model prediction results were compared and analyzed, while the MIC and common correlation coefficients: Pearson, Spearman, and Kendall correlation coefficients were used for variable screening and comparison, and the correlation coefficient (R 2 ), mean absolute percentage error (MAPE) as evaluation indexes, compared with the rest of the combined models, the MIC-LSTM model has the smallest deviation in the prediction results, the smallest number of feature variables in the public data set, and the highest MIC screening efficiency, which can improve the efficiency and effectiveness of nearshore significant wave height prediction.