Predicting Art Prices with Random Forest Regression

Art has always been more than just a form of expression. It has long since found its way into investor portfolios. Decisions around buying, selling or holding art have always been complex, not in any small part due to the notoriously opaque nature of the art market compared to traditional markets. Nevertheless, art is a resilient and potentially high-yielding asset class.

This projects attempts to provide an automated solution to certain challenges such as subjective valuation, and consequently, fluctuating prices. The data is Sotheby's auction data (retrieved from here: https://www.kaggle.com/code/ksesavinkova/sotheby-s-data-analysis/notebook ) and the resultant predictive model analyzes sales, artistic movements (like Baroque, Abstract, Surrealism, Post Impressionism, etc), year and period of creation, and condition factors.

Data was first preprocessed - the model cleans and transforms raw auction data and categorized conditions, periods, and signatures to extract any patterns that influence pricing. Next, I used a Random Forest Regressor to predict prices based on input features of year of creation, condition, artist signatures, and artistic period.

Included is a GUI that allows users to input artwork details and delivers a predicted price.

To access this, please download the project below (Python version 3.8).

ArtPricePredictor.zip

Download ZIP • 39KB