Supplementary Information for "Property Prediction and Explainable Analysis of Mn-Co-Ni NTC Thermistors by Direct Learning from XRD Patterns"

This dataset contains Python codes, input data, and output files used to investigate the relationship between X-ray diffraction (XRD) profiles and the electronic properties of Mn–Co–Ni thermistor materials using deep learning.train_baseline.py: Python script implementing convolutional neural network (CNN) models that directly learn from XRD profiles.train_baseline_with_descriptors.py: Python script implementing CNN models with compositional descriptors concatenated at the GAP (global average pooling) layer.00_dataset: Input datasets.XRD.csv: XRD patterns (CSV format).XRD_comp_desc.csv: Compositional descriptors combined with XRD patterns.01_outputs: Model outputs.*.pth files: optimized deep neural network parameters.*.csv files: predicted and measured values of thermistor properties (B constant and LogR).For further details, please refer to the paper: Yo Kato et al., Deep Learning–Based SEM Image Analysis for Predicting Thermistor Properties, DOI: 10.1080/27660400.2026.2637877