An Ensemble Machine Learning Approach to Estimating Swiss Home Prices

Sammanfattning: The price of a home is usually estimated by professional appraisers or by automated home valuation models, also known a hedonic pricing models. In the past, these models were largely based on linear regression models, however there is a trend towards using machine learning algorithms to estimate home prices. This thesis sheds light on automated home valuation models in Switzerland by developing a hedonic pricing model using machine learning algorithms and comparing it to the FPRE semiparametric regression model that is being used commercially to estimate home prices in Switzerland. The author found that a hedonic pricing model developed using a meta-learner trained on the predictions of several machine learning models was more accurate at estimating condominium prices in Switzerland than the FPRE model.