Meta-features for meta-learning

Meta-learning is increasingly used to support the recommendation of machine learning algorithms and their configurations. These recommendations are made based on meta-data, consisting performance evaluations characterizations prior datasets. characterizations, also called meta-features, describe properties data which predictive for trained them. Unfortunately, despite being in many studies, meta-features not uniformly described, organized computed, making empirical studies irreproducible hard compare. This paper aims deal with this by systematizing standardizing characterization measures classification datasets meta-learning. Moreover, it presents an extensive list tools, can be as a guide new practitioners. By identifying particularities subtle issues related measures, survey points out possible future directions that development meta-learning assume.