Academic Journal of Computing & Information Science, 2026, 9(8); doi: 10.25236/AJCIS.2026.090804.
Zhidong Zhao, Xindong Bai, Bo'en Zhao, Haoran Su, Sirui Zhao
Beijing 21st Century School, Beijing, China
Accurate used car valuation remains challenging, as transaction prices are jointly determined by multiple interacting factors including vehicle attributes, condition and market conditions, while raw listing data suffer from noise, incompleteness and severe right-skewed distribution. This paper proposes an end-to-end prediction pipeline for used car price estimation, which integrates robust data preprocessing, systematic feature engineering and a heterogeneous ensemble composed of four base models: LightGBM, XGBoost, CatBoost and a multilayer perceptron. Gradient-boosted tree models are trained on log-transformed targets with L1-type loss functions theoretically consistent with the mean absolute error (MAE) metric. Predictions from all base learners are aggregated via a non-negative weighted average, where the weights are optimized on out-of-fold predictions using the Nelder-Mead algorithm. Experimental results demonstrate that the ensemble model achieves superior predictive performance compared with individual models and traditional linear regression baselines. Error analysis indicates that residual errors are mainly concentrated in the sparse high-price vehicle segment, and the model implicitly captures duplicate vehicle records through fine-grained partitioning of embedding features. The findings verify that metric-aligned objective function design and the complementarity of heterogeneous learners, rather than the performance of any single model, are the primary drivers of the final predictive performance.
Used Car Price Prediction, Ensemble Learning, Gradient Boosting Decision Tree, Neural Network, Feature Engineering
Zhidong Zhao, Xindong Bai, Bo'en Zhao, Haoran Su, Sirui Zhao. A Heterogeneous Ensemble Learning Framework for Used Car Price Prediction. Academic Journal of Computing & Information Science (2026), Vol. 9, Issue 8: 28-35. https://doi.org/10.25236/AJCIS.2026.090804.
[1] Kim, M., Choi, J., Kim, J., Kim, W., Baek, Y., Bang, G., Son, K., Ryou, Y., & Kim, K. (2023). Trustworthy residual vehicle value prediction for auto finance. Proceedings of the AAAI Conference on Artificial Intelligence, 37(13), 15480–15488.
[2] Bukvic, L., Pasagic Skrinjar, J., Fratrovic, T., & Abramovic, B. (2022). Price prediction and classification of used-vehicles using supervised machine learning. Sustainability, 14(24), 17034.
[3] Amik, F. R., Lanard, A., Ismat, A., & Momen, S. (2021). Application of machine learning techniques to predict the price of pre-owned cars in Bangladesh. Information, 12(12), 514.
[4] Cui, B., Ye, Z., Zhao, H., Renqing, Z., Meng, L., & Yang, Y. (2022). Used car price prediction based on the iterative framework of XGBoost+LightGBM. Electronics, 11(18), 2932.
[5] Huang, J., Yu, Z., Ning, Z., & Hu, D. (2023). Used car price prediction analysis based on machine learning. In N. Radojević et al. (Eds.), Proceedings of the 2022 International Conference on Artificial Intelligence, Internet and Digital Economy (ICAID 2022) (pp. 356–364). Atlantis Press. https://doi.org/10.2991/978-94-6463-010-7_37
[6] Fayyaz, I., Ali, G. G. M. N., & Khairunnesa, S. S. (2025). Advanced feature engineering and machine learning techniques for high accurate price prediction of heterogeneous pre-own cars. Vehicles, 7(3), 94.
[7] Dutulescu, A. , Catruna, A. , Ruseti, S. , Iorga, D. , Ghita, V. , & Neagu, L. M. , et al. (2023). Car price quotes driven by data-comprehensive predictions grounded in deep learning techniques. Electronics, 12(14), 25.