Vehicle Price Prediction Using H2O-AutoML and Spark Based on Feature Engineering

In the automotive industry, the estimation of vehicle prices is critical for dealerships, consumers and manufacturers. Traditional methods struggle to capture the complexity of pricing factors. This paper proposes a framework for vehicle price prediction using big data technologies (H2O AutoML and Spark) and machine learning (ML) models. H2O AutoML automates model selection and tuning, and Spark facilitates efficient processing of large datasets. The proposed framework consisted of phase 1: EDA and Preprocessing, phase 2: Modeling, phase 3: Evaluation. The selected ML models are linear regression, Adaboost, Gradient Boosted Regression (GBR), random forest, and XGBoost, and are evaluated based on R2, Mean Squared Error (MSE), and Mean Absolute Error (MAE). The results indicate that the models consistently achieved higher R-squared (R2) values compared to benchmark studies with the best performing model being GBR, achieving an R2 value of 98.1% and exceeding previously reported values of 90%. These results suggest that manufacturers can use this framework to set competitive prices. Dealerships can then optimize their pricing strategies, and consumers can make more informed purchasing decisions. Big data and machine learning techniques have the potential to revolutionize vehicle pricing strategies and enhance the car-buying experience.