Using meta-learning for automated algorithms selection and configuration: an experimental framework for industrial big data