A Ridge Regression-Based Hybrid Model for Stock Index Forecasting

This study introduces an innovative stock index forecasting method that combines Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (CEEMDAN) and Ridge Regression (RR) algorithms. CEEMDAN adaptively decomposes complex financial time series into intrinsic mode functions (IMFs) and a residual, capturing market volatility at various frequencies. Ridge Regression is then applied to the decomposed components to address multicollinearity, improving the stability and accuracy of predictions. Experimental results demonstrate that the CEEMDAN-RR model outperforms traditional linear and nonlinear models, showing strong fitting ability in short-term predictions. Furthermore, it is more robust and accurate than deep neural networks in managing market nonlinearities and noise.