SHAPCA: Consistent and Interpretable Explanations for Machine Learning Models on Spectroscopy Data
Mingxing Zhang, Nicola Rossberg, Simone Innocente, et al.
This paper presents SHAPCA, a method for explaining machine learning predictions on spectroscopy data (like chemical or medical scans). The approach combines dimensionality reduction and a popular explanation technique called SHAP to provide interpretable results that connect back to the original spectral signals, rather than abstract processed features. The method shows better consistency and stability in explanations compared to existing approaches, making it more trustworthy for clinical and safety-critical applications.
explainabilityinterpretabilitySHAPPCA