A fast prediction method for sea surface scattering coefficient based on principal component analysis dimensionality reduction

The electromagnetic scattering characteristics of sea surface have complex dependencies on various influencing factors, such as sea wave parameters, radar parameters, etc. The traditional electromagnetic scattering prediction models for the large-scale sea surface tend to suffer from overfitting problems when facing multi-parameter high-dimensional mapping. Choosing appropriate dimensionality reduction methods and model parameters is an efficient way to improve the model performance. Therefore, this paper proposes a fast prediction method for sea surface electromagnetic scattering based on principal component analysis dimensionality reduction. Firstly, the backscattering coefficient data set is constructed by using the Wen’s spectrum and multi-scale electromagnetic scattering model of the sea surface. Then, the principal component analysis method is introduced to reduce the dimension of the simulation parameters and extract the main features. Finally, a nonlinear regression model based on least squares support vector regression(LSSVR) machine is established, and the dimensionality reduction data is inputted for prediction and the accuracy of the prediction results is evaluated. By comparing the prediction results of different dimensionality reduction ratios, the influence of principal component dimensionality reduction on the model performance is analyzed. The results show that reducing the dimension of the simulation parameters appropriately can significantly increase the accuracy and enhance the interpretability. When the dimensionality reduction ratio is about 25%, the model accuracy reaches the optimum. When the dimensionality reduction ratio is greater than 40%, the model accuracy decreases significantly, which is not conducive to sea surface electromagnetic scattering prediction.