Improving accuracy models using elastic net regression approach based on empirical mode decomposition

In this study, an elastic net (EN) regression model based on the empirical mode decomposition (EMD) algorithm is used in two applications, namely, numerical experiment and actual time series data. EMD is used to analyze a nonstationary and nonlinear signal dataset, which includes a set of orthogonal intrinsic mode functions (IMFs) and residual components. EN regression is used to select the most significant predictor variables influencing response variables and can address the multicollinearity problem between predictor variables. The main objective of this study is to apply the proposed method, EMD-EN, by using two variables for selecting important orthogonal IMFs and the residual components of predictor variables with significant effects on response variables. Moreover, this study uses the EMD-EN method in two different applications involving nonstationary and nonlinear problems. Results show that the proposed method outperforms other competitive methods in the numerical experiment and applications.