Prediction of lithium-ion conductivity in solid-state electrolytes by combined use of physical descriptors and element embeddings

The file DDSE_TRAIN.csv contains the full dataset used for model training. The file DDSE_TEST_RESULTS.csv includes both the experimental (target) and model-predicted ionic conductivity values used for testing. Additionally, external validation datasets containing 657 data points are provided as DATASET_A_VALIDATION_RESULTS.csv, DATASET_B_VALIDATION_RESULTS.csv, and DATASET_C_VALIDATION_RESULTS.csv. In these validation files, log10_target represents the experimentally reported ionic conductivity values, while log10_predict represents the ionic conductivity predicted by the trained machine learning model.