Optimizing Modelling Accuracy Using Variational Mode Decomposition and Elastic Net Regression: Evidence in Stock Market Prediction

Accurate modelling of complex, nonlinear and nonstationary datasets remains a critical challenge in predictive analytics. This study introduces a novel variational mode decomposition-elastic net regression (VMD-Enet) framework that combines VMD with ENet to enhance prediction accuracy and interpretability. VMD first decomposes signals into intrinsic mode functions (IMFs), effectively denoising data and improving feature representation. ENet is then applied to select the most significant predictors while managing multicollinearity. The proposed approaches are evaluated using numerical simulations and real stock market data. The proposed VMD-ENet model demonstrated superior performance over the other methods. In the case of the stock market experimental analysis, VMD-ENet achieved the lowest errors, with RSS = 88.90, RMSE = 0.837, MAE = 0.668, and WQE = 0.0006. Compared to other regularization approaches, VMD-ENet significantly identifies key predictors without arbitrarily discarding correlated variables, ensuring model stability. These findings highlight the framework's robustness, interpretability and predictive superiority, making it a promising tool for financial market analysis and broader applications in complex data modelling.