A Summer Precipitation FNN Multi-step Prediction Model Based on SSA-MGF

A fuzzy neural network (FNN) multi-step prediction model based on singular spectrum analysis (SSA) and mean generating function (MGF) for summer precipitation has been developed in this paper. In the modeling process, the original standardized sample series of summer precipitation was denoised and reconstructed with SSA, the extended matrix of MGF of the reconstructed precipitation series (as the input factor) and the original standardized sample series (as the output factor) were then used to develop a three-layer FNN multi-step prediction model for summer precipitation. Results show that the SSA-MGF FNN model is superior to the other three models in prediction accuracy. This indicates that denoising of SSA and FNN prediction model are relatively effective for raising the accuracy of precipitation prediction, and the SSA-MGF FNN multi-step prediction model proposed in this paper is of application value.