What is the Price of Your Used Car? Automated Predictions using XGBoost and Neural Networks

As the used car market presents a high economic relevance, the importance of accurately predicting prices to enable informed decision-making for buyers and sellers has become more pressing. Experiments with Machine Learning methods have been made to create robust prediction models. In this study, after analyzing various state-of-the-art methods, we introduce a dataset of over 25,000 entries from the Romanian market and experiment with different approaches involving extreme gradient boosting and deep learning. We propose a neural network architecture that captures complex relationships between the car model particularities and individual add-ons. Our findings reveal that the proposed approach accurately predicts used car prices with a mean absolute percentage error of 10.68% while providing insights into the significant factors influencing the price.