Deriving equation from data via knowledge discovery and machine learning: A study of Young’s modulus of Ti-Nb alloys

[Display omitted] •A property-targeted machine learning regression model can discover knowledge inside materials datasets and facilitate the machine learning process.•A framework of unearthing inner relationships is summarized, where domain knowledge runs through and machine learning provides insights.•An accurate, interpretable, and rational equation for Young’s modulus of Ti-Nb alloys is found, analyzed, and validated based on a customized workflow (come from the proposed framework). Acquiring knowledge and assisting materials design from computing and experimental data is very interesting and important at the intersection of materials and data science. In this work, we construct a whole framework to find an easy-to-interpret equation by taking a study of Young’s modulus of Ti-Nb alloys as an example. Here, we used Young’s modulus-targeted regression model (decision tree) to find a potential rule for classifying the dataset. The explicit equation was constructed through machine learning (ML) and validated based on physical laws. The transferability of the equation stood out after comparing it with five ML models including support vector machine (SVR), linear regression (LR), k-nearest neighbor (KNN), decision tree (DT), and random forest (RF). The results prove that our method indeed helps us to discover the contained knowledge hidden in the data and to uncover the relationship between features and property.