Diamond Price Prediction Based on Regression Algorithms: Analyzing Key Factors and Model Performance

This research investigates the application of machine learning approaches for forecasting diamond prices and determining the main factors that influence pricing within the jewelry sector. The dataset used comprised 53,940 entries, each with 10 attributes. Data preprocessing involved eliminating invalid entries and encoding categorical attributes. Various regression models, including random forest, k-nearest neighbors, Extreme Gradient Boosting (XGBoost), and multilayer perceptron, were assessed. The cross-validation results revealed that the XGBoost model achieved the best performance, delivering the lowest Root Mean Square Error (RMSE) and high predictive accuracy. In testing, the XGBoost model obtained an R-squared (R) of 0.982129 and an adjusted R of 0.982114. The critical factors affecting diamond prices were identified as carat, cut, color, and clarity. This study offers a systematic framework for pricing strategies in the jewelry market and emphasizes the significant role of machine learning in predictive modeling. Further research should aim to refine feature engineering and data processing to enhance accuracy and expand its practical application.