Advances in International Applied Mathematics

Accurate precipitation prediction has a great significance to the prevent and control of rainstorm and flood disasters. Due to the lack of information mining ability of traditional precipitation prediction methods, the improved Variational Mode Decomposition (VMD) and Support Vector Regression (VMD-SVR) prediction method was proposed. Taking Zhengzhou as an example, firstly the chaotic characteristics of monthly precipitation series in the city from 1979 to 2020 were analyzed. Secondly, the original series data, phase space reconstruction data and variational mode decomposition data of precipitation series are predicted. The results show that the data prediction performance after variational mode decomposition is higher. Finally, the prediction accuracy of the three machine learning algorithms was compared, the VMD-SVR model was found to have the highest prediction accuracy.

Variational Mode Decomposition; Support Vector Regression; Chaotic Time Series; Prediction of precipitation