The paper introduces a data-driven and model-free framework called Feature-and-Reconstructed Manifold Mapping (FRMM) to predict multiple components in complex systems. FRMM combines feature embedding and delay embedding to find low-dimensional manifolds that are topologically equivalent to the high-dimensional dynamical system. These low-dimensional manifolds serve as generalized predictors for all components, overcoming the curse of dimensionality. The framework is validated on both benchmark models (Lorenz and Rössler systems) and real-world datasets (Indian monsoon, EEG signals, foreign exchange market, and traffic speed in Los Angeles). FRMM demonstrates reliable predictions with high accuracy and robustness, outperforming traditional methods and other STI-based frameworks. The key advantages of FRMM include its ability to handle high-dimensional systems, interpretability, and potential applications in various fields.The paper introduces a data-driven and model-free framework called Feature-and-Reconstructed Manifold Mapping (FRMM) to predict multiple components in complex systems. FRMM combines feature embedding and delay embedding to find low-dimensional manifolds that are topologically equivalent to the high-dimensional dynamical system. These low-dimensional manifolds serve as generalized predictors for all components, overcoming the curse of dimensionality. The framework is validated on both benchmark models (Lorenz and Rössler systems) and real-world datasets (Indian monsoon, EEG signals, foreign exchange market, and traffic speed in Los Angeles). FRMM demonstrates reliable predictions with high accuracy and robustness, outperforming traditional methods and other STI-based frameworks. The key advantages of FRMM include its ability to handle high-dimensional systems, interpretability, and potential applications in various fields.