Machine Learning Choices for Diamond Price Prediction, A Controlled Empirical Study

DOI

10.2991/978-94-6239-801-6_142How to use a DOI?

Keywords

Diamond price prediction; tabular regression; supervised learning; gradient boosting; error analysis

Abstract

Diamond price prediction is an attribute-based regression problem in which observable diamond characteristics explain a substantial but incomplete part of market price. This paper converts a public Diamonds-dataset baseline into a controlled empirical study of modelling choices. The workflow reproduces the baseline preprocessing and evaluates five hypotheses concerning log price transformation, volume feature engineering, categorical encoding, model family, and simple ensembling. RMSE is used as the primary metric, with MAE, R-squared, and cross-validation RMSE as supporting metrics. Results show that tree-based and boosting models outperform linear regression and support vector regression. The best individual extended model is CatBoost, while a simple ensemble of Random Forest, XGBoost, LightGBM, and CatBoost achieves the lowest test RMSE, reducing RMSE from 554.73 for the reproduced XGBoost baseline to 521.16. Segment-level diagnostics show broadly similar relative MAE across low-, mid-, and high-price groups, but dollar-denominated errors increase for high-price diamonds and repeated splits show mild high-price underestimation. These findings indicate that attribute-only valuation is accurate on average but remains limited in the upper price range, where richer market, certification, and transaction variables are likely needed.

Copyright

© 2026 The Author(s)

Open Access

Open Access This chapter is licensed under the terms of the Creative Commons Attribution-NonCommercial 4.0 International License (http://creativecommons.org/licenses/by-nc/4.0/), which permits any noncommercial use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license and indicate if changes were made.

Download article (PDF)

TY  - CONF
AU  - Zekai Hu
PY  - 2026
DA  - 2026/10/23
TI  - Machine Learning Choices for Diamond Price Prediction, A Controlled Empirical Study
BT  - Proceedings of the 2026 2nd International Conference on Financial Innovation and Marketing Management (FIMM 2026)
PB  - Atlantis Press
SP  - 1366
EP  - 1376
SN  - 2352-5428
UR  - https://doi.org/10.2991/978-94-6239-801-6_142
DO  - 10.2991/978-94-6239-801-6_142
ID  - Hu2026
ER  -