SAE-extreme random tree-based wave forecasting method

The invention provides a wave forecasting method based on an SAE-extreme random tree, and the method comprises the steps: firstly collecting the historical significant wave height, wind speed and wind direction data of a to-be-predicted point, predicting + 22h and + 24h wind speeds through employing a seasonal difference moving average autoregression model, and then employing the-71h-0h significant wave height, wind speed and wind direction time sequence and + 22h and + 24h wind speed prediction values as features, employing the + 24h significant wave height as a target value, and carrying out the prediction of the + 22h and + 24h wind speeds. The method comprises the following steps: establishing a sample data set, finally training a sparse auto-encoder on the data set, extracting a sparse auto-encoder (SAE) hidden layer, and training an extreme random tree together with the original data set, thereby obtaining a complete + 24h significant wave height prediction model, and solving the problem that the freque