Fig. 5
From: Predicting refractive index of inorganic compounds using machine learning
(a) The top 20 features that impact the prediction of the RFR, arranged from highest to lowest importance. The horizontal axis shows SHAP values, indicating each feature impact on predictions. Features are ranked by importance on the vertical axis. Dots represent data instances, colored from blue (low) to red (high). (b) MAE of RFR, GBTR and ERTR methods for 10-fold cross-validation. \(\mathcal {D}_1\), \(\mathcal {D}_2\), \(\mathcal {D}_3\), and \(\mathcal {D}_4\) represent data sets with 1, 5, 21, and 10 predictors, respectively. (c) Refractive index (RI) predicted using empirical formulas and prediction method of this study(ERTR\(\mathcal {D}_4\)). The solid black line indicates \(x=y\).
