Phase Classification of Multi-Principal Element Alloys via Interpretable Machine Learning

This dataset contains 1821 compositions ranging from binary to multi-component alloys along with phase information by referring to several previous reports that compiled experimental data from the published literature.A total of 125 variables for each observation were generated by the Magpie program and down-selected to 12 variables based on linear Pearson correlation coefficient and non-linear normalized mutual information analyses. The down-selected variables can be subdivided into three categories: (1) those that are chemistry-agnostic (e.g., MixingEntropy),(2) those that depend on element pairs (e.g., DeltaHf), (3) those that are depend on chemistry (e.g., maxdiff_Electronegativity).

本数据集涵盖1821组从二元到多组分合金的成分数据，并附带相信息，所有数据均参考了多篇从已发表文献中整理实验数据的既往研究成果。研究通过Magpie程序为每条样本生成了共计125个特征变量，随后基于线性Pearson相关系数与非线性归一化互信息分析，将特征筛选至12个。经筛选得到的特征可划分为三大类别：(1) 化学组分无关型特征（例如混合熵MixingEntropy）；(2) 元素对依赖型特征（例如DeltaHf）；(3) 化学组分依赖型特征（例如电负性最大差值maxdiff_Electronegativity）。