Revisiting Identity and Spectra Dispersion in Media-Bridged Time Series Forecasting: Linking Multivariate Signals and Narrative Flows
The research introduced MIDAPN, a unified forecasting backbone for media-bridged time series data combining numerical and narrative flows. It addresses identity entanglement by using a graph to analyze media essence and behavior, and solves temporal scale mismatch with Spectral Prism Convolution. The model proves superior across diverse datasets, establishing a robust method for linking different types of time series signals.