Air Temperature Prediction Based on EMD and LS-SVM

Abstract

Air temperature is closely related to life and affects all aspects of life. Therefore, the forecast of the temperature is more far-reaching. In this paper, a new model based on EMD (Empirical Mode Decomposition) and LS-SVM (Least Squares Support Vector Machine) was proposed. At first, EMD was applied to adaptively decomposing the time series into a series of different scales of intrinsic mode function. Then, for each intrinsic mode function, using the appropriate kernel function and model parameters construct different LS-SVM to predict the temperature. Finally, the predicted values of each component were fitted to get the final forecast. Compared with the single LS-SVM and neural network prediction method, simulation results showed that the method in this paper has higher accuracy.