Predicting housing sale prices in Germany by application of machine learning models and methods of data exploration

The prediction of real estate prices is a popular problem in the field of machine learning and often demonstrated in literature. In contrast to other approaches, which regularly focus on the US market, this paper investigates the biggest, German real estate dataset, with more than 1.5 million unique samples and more than 20 features. In this paper we implement and compare different machine learning models in respect to performance and interpretability to give insight in the most important properties, which contribute to the sale price. Our experiments suggest that the prediction of sale prices in a realworld scenario is achievable yet limited by the quality of data rather than quantity. The results show promising prediction scores but are also heavily dependent on the location, which leaves room for further evaluation.