The purpose of this thesis is to investigate the accuracy of different estimation and forecasting methods for residentialreal estate prices. Therefore, we use four different statistical methods - a multivariate linear regression, generalized linear model, generalized additive model and gradient boosting decision tree. The basis of the investigation is a data setconsisting of 3918 residential property price observations in Munich, Germany. The estimation performance of thesemethods is tested by cross validation. The more sophisticated the regression, the stronger is the improvement of the results. The generalized additive model and gradient boosting have similar results with the latter one being slightlybetter. The forecasting task sheds a different light on the methods. In order to test the performance, we fit the models on a data set containing residential property price observations from 2013 to 2017 and test the models with theobservations from 2018. In this regard, the generalized additive model clearly outperforms the others, whereas especiallythe gradient boosting tree is not competitive.     «

The purpose of this thesis is to investigate the accuracy of different estimation and forecasting methods for residentialreal estate prices. Therefore, we use four different statistical methods - a multivariate linear regression, generalized linear model, generalized additive model and gradient boosting decision tree. The basis of the investigation is a data setconsisting of 3918 residential property price observations in Munich, Germany. The estimation performance of thesemethods is tested by c...     »