Diamond Price Prediction Using Data Mining Techniques

Abstract

Data mining is the technique of uncovering different patterns and finding relationships and anomalies in the large datasets. It is used for the extraction of valuable information from the available data present in the form of a dataset. Prediction has been proposed as one of the useful techniques of data mining which can be used for predicting about future trends. Generally, regression is preferred for prediction. In a regression problem, the final output is calculated as the mean of all outputs, which is also called Aggregation. In this paper, a diamonds dataset was downloaded from Kaggle. The dataset consists of 53,940 instances and 10 features. The following dimensions were used for predicting the prices of a diamond: carat, cut, color, clarity, depth, table, price, x, y and z. Based on the above, three regression algorithms, i.e., Random Forest Regressor, K-Neighbors Regressor and Decision Tree Regressor and one optimization technique, i.e., Particle Swarm Optimization had been applied and the performance of the four algorithms were compared based on different performance and validation metrics such as testing accuracy (R2-score), Mean Absolute Error (MAE) and Root Mean Squared Error (RMSE). The testing accuracy of Random Forest Regressor was found to be 0.9822175620286142, Particle Swarm Optimization 0.9809447466912986, K-Neighbors Regressor 0.9683118493918951 and Decision Tree Regressor 0.9675318135094464. It was observed that the testing accuracy of Random Forest Regressor was the best as compared to the rest others.