Cut diamonds are hard to value given the number and type of properties used in price construction. This project aims to develop a valuation model for cut diamonds based on data published on the Internet. Regression trees (Classification and Regression Trees and Chi‐Square Automatic Interaction Detection) and neural networks (using backpropagation) are used for this purpose. The proposed approaches have a complementary role in the application. Neural networks have a better performance in prediction, accounting for around 96% of cut diamond unit prices variation. The role of regression trees is fundamental in interpretability, helping to understand the contribution of predictors in pricing. The models' results may prove to have some advantages over the Rapaport price lists (an industry‐wide adopted price indicator).

--

Prediction,Data mining,Regression trees,Neural networks,Cut diamond price

• Mathematics - Natural Sciences
• Computer and Information Sciences - Natural Sciences
• Economics and Business - Social Sciences
• Other Social Sciences - Social Sciences