Machine learning and human analysts in predicting auction prices of Artworks

Abstract

This thesis aims to research the factors that influence the final prices of artworks in the auction market. To achieve this, a dataset is constructed by web scraping data from Christie's website. Machine learning techniques are applied to process the data and to identify the variables that affect the final price of an artwork. At the end a predictive model is individuated and compared with a benchmark that involves the forecasts made by experts. The ultimate goal is to enhance the understanding of the art market pricing mechanisms and provide valuable forecasting abilities by minimizing the subjectivity often present in the industry through a data-driven approach.