ML-Driven Valuation of Used Smartphones

Abstract

In the current digital age, the rapid pace of smartphone upgrades complicates the accurate estimation of resale values in the second-hand market. The primary objective of this research is to develop an efficient machine learning–based system for forecasting the resale prices of pre-owned smartphones and to evaluate the effectiveness of multiple regression models. A manually collected dataset comprising more than 900 smartphone records was compiled from online marketplaces and local mobile shops to ensure realistic and up-to-date pricing information. Regression techniques, including Linear Regression, Decision Tree, Random Forest, Gradient Boosting, and XGBoost, were implemented and analyzed using standard performance metrics such as R², mean absolute error (MAE), and root mean square error (RMSE).. The comparative analysis identifies the most suitable model for accurate price prediction, with ensemble-based methods demonstrating superior performance over traditional approaches. The proposed methodology provides a reliable and data-driven framework that enhances pricing transparency, reduces reliance on subjective retailer assessments, and supports informed decision-making for both buyers and sellers in the used smartphone resale market.