Diamond Price Prediction using Machine Learning
Harshvadan Mihir, Manish I. Patel, Soham Jani, Ruchi Gajjar
Abstract
Diamond is one of the strongest and the most valuable substances produced naturally as a form of carbon. However, unlike gold and silver, determining the price of a diamond is very complex because many features are to be considered for determining its price. This paper aims to come up with the most efficient algorithm for the price prediction of diamonds. The algorithms such as Linear regression, Support Vector regression, Decision trees, Random Forest regression, K-Neighbors regression, CatBoost regression, Huber regression, Extra tree regression, Passive Aggressive regression, Bayesian Regression and XGBoost Regression are used to train the particular machine learning models on the diamond dataset for the prediction of diamond prices based on various attributes. The comparative analysis of various Machine Learning Regression models is done for the price prediction of any diamond. From the performance parameter values and analysis, it was found that the CatBoost Regression algorithm proved to be the most optimal algorithm having an R2 score of 0.9872 and formidable training and testing accuracies of 98.74% and 98.72% respectively. Hence, the CatBoost algorithm has been implemented for the price prediction of a diamond specimen with the help of the values of attributes extracted from an image of a diamond certificate.