Fig. 5: Machine learning-assisted discovery of descriptors for electrode kinetic matching. | Nature Communications

Fig. 5: Machine learning-assisted discovery of descriptors for electrode kinetic matching.

From: Machine learning-assisted kinetic matching model for rational electrode design in aqueous zinc-ion batteries

Fig. 5

a An overview of ML-based model. b Heatmap analysis between all features, including average diffusion coefficients of positive electrode (+D) and negative electrode (−D), standard deviation of positive electrode (+V) and negative electrode (−V), DR, Sum of standard deviation (S), overlap_index (OI), CSI, weighted diffusion discrepancy (WD), Asymmetry Index (AI), RC, stability (Sta). c SHAP values of feature importance. d Order of relative contribution among variables. e Modeling the performance of features for single-target prediction using the XGBoost model. f The predicted value numerical range corresponding to the DR and CSI ranges, the blue solid line represents all predicted sample values, the orange dashed line indicates the values with a performance threshold of top 30%, and the gray dashed lines demarcate the feature value range corresponding to the top 30% density distribution.

Back to article page