Dual-Modality Machine Learning: Enhancing Predictions with NIR Spectra and Interferogram Data Fusion

Within the realm of analytical measurements and machine learning (ML), the fusion of diverse data modalities opens avenues for significantly improving prediction models. In the context of Near-Infrared Spectroscopy (NIRS), the conventional approach has predominantly relied upon Fourier transform spectral data as the foundation for predictive model construction. Building on this established foundation, this study introduces a novel dual-model framework that harnesses the advantages of fusing information from both the spatial interferogram data and the Fourier transform spectral data, which are then incorporated into a machine learning stacked architecture. This approach leverages the high-dimensional spectral information provided by the Fourier transform spectral data and the unique spatial patterns captured by interferograms, facilitating a comprehensive analysis that surpasses the capabilities of models based on a single data domain. The findings of the study demonstrate a remarkable improvement in model performance, characterized by a notable 34% reduction in Root Mean Square Error (RMSE) alongside a substantial 13.4% increase in the coefficient of determination (R²). These results emphasize the transformative potential of the dual-model framework in advancing predictive modeling and unlocking deeper insights into the complex interplay of spatial and spectral data characteristics contributing to advancements in analytical precision and predictive modeling within the NIRs domain.