Do prices of artworks result from the emotions of art lovers or are there explanatory, natural factors? The goal of this project was to develop a model that can automatically estimate auction prices of artworks using past auction data and information about the art market. Machine learning algorithms were used to train models to predict the auction price of an artwork.

The results showed that this allowed automated auction prices to be estimated similarly to those estimated by auction house experts. Indeed, the prices can be explained primarily by past auction prices of the respective artist and less by picture characteristics. The developed model can be integrated into an application that supports auction houses, galleries and art dealers in art valuation.

With:

• Thomas Maier, Senior Expert Data Scientist, Datahouse
• Andreas Thürlimann, Data Scientist, Datahouse
• Johann Burkhard, ArteMorfosis
• Bruno Y. Thalmann, Art Leasing & Invest AG
• Leo Herrmann, Corporate Communications (ETH Zürich), Moderation