The aim of this work was to evaluate the applicability of models based on artificial intelligence (exactly machine learning) with respect to the prediction of scrap-tolerant aluminum alloys with a particular focus on AlMgSi alloys (6xxx). For this purpose, a structured data collection based on literature values was created. From this collection of 271 data points, 221 were selected and submitted to Citrine Informatics. There, the data have been used successfully for generating two machine learning models. The two models, based on the chemistry of the alloy, can predict (i) tensile strength (ii) elongation to failure. By using the models, predictions for 2280 chemical compositions could be generated. Parallel to the model development, the existing data collection was extended experimentally. First, a detailed analysis of the already collected data took place, followed by the determination of the chemical compositions of the alloys to be produced. Particular attention was paid to the scrap tolerance of the produced materials. Before casting, the process and heat treatment temperatures were determined by thermodynamic calculations. 14 alloys with varying contents of Mg, Si, Zn and Cu were successfully cast and processed to sheet material. Tensile tests were utilized to determine the mechanical properties of the materials in the T4-state. In addition, the as-cast and solution-annealed states of each produced alloy were examined using scanning electron microscopy. The results proved that even high-alloyed, scrap-tolerant aluminum alloys can achieve very good mechanical properties. In addition, the data obtained from the experiments significantly extended the range covered by the data collection. With the additional experimental data, the total data collection contains 293 data points at the time of completion of this work.