Research on Visualization Techniques Based on High-Dimensional Attribute Data of Chemical Materials

Machine learning is currently used to analyze and predict materials in fully automated applications in the materials domain. However, human interpretation and involvement are limited due to the opaque nature of these algorithms. This study explores for the first time the use of semi-human and semi-automated analysis techniques in materials data research. The analytics approach combines machine learning techniques with data visualization and analysis so that the analysis of material properties is dominated by human intelligence, with machine learning techniques as a complementary method. This helps experts understand the associations between material properties from a broader perspective, enabling them to generalize the concept of micro-properties to macro-properties.