Machine learning has become increasingly important in recent years and is increasingly penetrating a wide variety of application ￿elds. In particular, machine learning-based methods have been used in the past to calculate approximate solutions where classic algorithms have failed. However, to the best of the authors’ knowledge, the authors have examined the properties of machine learning models generated from data that control classic calculation processes. As part of this work, we use the example of comparison sorts to investigate the properties of such models generated from data. For this purpose, we examine comparison sorts in regard to their building blocks and de￿ne a model of of computation as a theoretical basis. Based on this, we implement a representative selection of comparative sorts and evaluate properties of the training data generated in this way.