Predictive modeling for pricing used vehicles in the Jordanian market: A machine learning approach

The Jordanian automotive market is expected to grow and reach a value of US$3 billion. Although car prices in Jordan are higher compared to neighboring regions, there has been significant growth in car sales. However, a lack of market understanding leads to exploitation of buyers and sellers. Machine learning models have proven effective in analyzing data patterns and making accurate predictions when designed properly. This research utilizes machine learning algorithms and data processing techniques to predict the prices of used cars in the Jordanian vehicle market. The study has three key contributions. Firstly, it involves creating a dataset with information on used vehicles in the Jordanian market. Secondly, it includes data preprocessing and the application of feature engineering techniques to improve the dataset. Lastly, the study applies several machine learning approaches, specifically Random Forest and XGBoost, known for their robustness and accuracy in similar predictive tasks. Our experiments show that the Random Forest model excels in classification tasks with an F1 Score of around 0.81, demonstrating its effectiveness in handling imbalanced data typical in automotive pricing. Additionally, XGBoost is favored for regression tasks, achieving an R-squared score of approximately 0.87, outperforming traditional linear regression methods. These models can greatly benefit both buyers and sellers by predicting appropriate car prices in the Jordanian used car market, instilling confidence, and reducing the risk of exploitation.