Predicting and Interpreting the Spectroscopic Properties of Nd·Y Co-doped Fluorites with Machine Learning

1/ AI 활용 방법: linear regression, support vector regression, BPNN, XGBoost, and SHAP interpretation
2/ 적용 과제: prediction and interpretation of spectroscopic properties of Nd/Y co-doped fluorites from local coordination structures
3/ 보고된 결과: structure-property correlations for absorption, emission, and lifetime were quantified, and a data-driven optimization path was proposed
4/ npj Computational Materials reported a machine-learning study that predicts and interprets spectroscopic properties from local coordination environments in an Nd and Y co-doped fluorite platform.
5/ The paper stated that linear regression, SVR, BPNN, and XGBoost were applied, and SHAP was used to interpret variable contributions, quantifying structure-property relationships for absorption, emission, and lifetime.
6/ From a materials research perspective, this work is a data-driven case showing that spectroscopic properties can be more sensitive to local structure than to simple composition alone.
7/ It also demonstrates a way to complement optimization problems that are difficult to handle with DFT alone by combining interpretable regression and boosting models with explanation methods.
8/ However, these results are based on a specific Nd/Y co-doped fluorite dataset and model combination, and they cannot be directly generalized to other compositions or other spectroscopic systems.
9/ Prediction performance and interpretation depend strongly on the input structural descriptors, dataset size, and distribution range, so the applicable scope must be checked. [1]