Using Physical Parameters for Phase Prediction of Multi-Component Alloys by the Help of TensorFlow Machine Learning with Limited DataUsing Physical Parameters for Phase Prediction of Multi-Component Alloys by the Help of TensorFlow Machine Learning with Limited Data

In recent years developing new material and compounds have become more important because of the community’s needs. Material scientist and physicist great effort make significant changes in daily life. But nowadays it is important to make these changes in a short time. In this point of view, artificial intelligence and machine learning gives the scientist a great opportunity to predict the properties of new compounds before produced in the laboratory. In this study, the valence electron concentration (VEC), atomic size difference (δ), enthalpy of mixing (∆H_mix), the entropy of mixing (〖∆S〗_mix) and electronegativity difference (∆χ) values are calculated for each alloy and a dataset has been created. We use gradient boosted trees machine learning method with TensorFlow artificial intelligence program to explore phase selection using an experimental dataset consisting of 118 multi-component alloy system. We divide the whole dataset into two portions with training and evaluate dataset. The training dataset contains 73 and evaluate dataset contains 45 multi-component alloy systems. We also show three of the predicted multi-component alloy system to examine which physical values are used predominantly during prediction. We look at the Receiver Operating Characteristic (ROC) of the results, which will give us a better idea of the tradeoff between the true positive rate and false positive rate. It has been observed that this learning method predicts the structure correctly in 95% of the results with limited data.In recent years developing new material and compounds have become more important because of the community’s needs. Material scientist and physicist great effort make significant changes in daily life. But nowadays it is important to make these changes in a short time. In this point of view, artificial intelligence and machine learning gives the scientist a great opportunity to predict the properties of new compounds before produced in the laboratory. In this study, the valence electron concentration (VEC), atomic size difference (δ), enthalpy of mixing (∆H_mix), the entropy of mixing (〖∆S〗_mix) and electronegativity difference (∆χ) values are calculated for each alloy and a dataset has been created. We use gradient boosted trees machine learning method with TensorFlow artificial intelligence program to explore phase selection using an experimental dataset consisting of 118 multi-component alloy system. We divide the whole dataset into two portions with tr