RVFL-NLNB Rainfall Forecasting Model Based on Feature Extraction of MGF and PCA

Abstract

In view of the difficulty in determining the modeling factors and the poor stability of the prediction model in precipitation prediction modeling. In this paper, firstly, the extension factor matrix is generated by using the mean generation function, and the dimension of the extension matrix is reduced by making use of the principal component analysis technology. The effective data features are extracted as independent variables and the original precipitation series as dependent variables. Secondly, by comparing the architecture of random vector functional link model and random vector functional link that has no input-to-output links and no bias in output neurons, then the random vector functional link with no input-to-output links and no bias in output neurons prediction model of monthly precipitation in Liuzhou city is established. Finally, the prediction model is established for Liuzhou April precipitation data. This method makes full use of the reconstructed data of mean generation function and principal component analysis to produce precipitation factors, and employs random vector functional link with no input-to-output links and no bias in output neurons network model, the monthly precipitation prediction model of Liuzhou is established. The experimental results express that the model has reliability, which provides a reliable method for precipitation prediction.