Prédiction des propriétés des matériaux par apprentissage automatique.

Abstract

In the last years, the materials science community has made considerable efforts to use informatics to accelerate the development and discovery of new materials. The algorithms of machine learning analyze material properties data to extract new knowledge or to predictive models representing the behavior of materials from existing databases in materials science. This technique is less expensive in computing time than traditional ab-initio codes. In this master’s thesis, we have implemented the algorithms of machine learning, in python, using Scikit-learn to extract data from platforms such as Materials Project and Citrination.