Multicellular Gene Expression Programming-Based Hybrid Model for Precipitation Prediction Coupled with EMD

Abstract

Accurate and timely precipitation prediction is very important to people’s daily activities and production plans. However, the impact factors of meteorological precipitation are numerous and complex, making it difficult to predict, and the prediction effect by traditional methods is difficult to meet the public expectations. This work proposes to use Multicellular Gene Expression Programming (MC_GEP) algorithm for modeling the historical precipitation data series decomposed by Empirical Mode Decomposition (EMD). Then we design a novel Multicellular Gene Expression Programming based method coupled with Empirical Mode Decomposition, named as EMGEP2RP, for precipitation modeling and prediction. Using RMSE and MAE as evaluation indicators, simulation experiments were conducted on three different types of real precipitation data sets in different regions. The comparing results show that the EMGEP2RP algorithm significantly outperforms not only the existing Gene Expression Programming (GEP) algorithm, but also the Back Propagation and Support Vector Machine algorithms which are widely used in meteorological modeling and predictions.