Research on Used Car Price Prediction Based on Stacking Model Fusion

With the continuous increase in the number of motor vehicles, the per capita ownership also increases, and the demand for second-hand car circulation is increasing, but the price of second-hand cars is difficult to accurately estimate and set, which is not only affected by the basic configuration of the car itself, but also by factors such as the condition of the car. This article will help used car trading platforms to solve this problem through data analysis and modeling, and build a machine learning model using a large amount of historical used car transaction data, and analyze the modeling process in detail from data preprocessing, feature engineering, and modeling parameter adjustment. Three machine learning models with good performance in regression problems, namely gradient lifting decision tree GBRT, ultimate gradient lifting algorithm XGBoost and light gradient lifting machine LightGBM, were fused by Stacking model fusion method. The simulation data shows that the average absolute error of the Stacking model fusion is reduced by 2.7% compared with the single model with the best prediction performance, which can be used to solve the practical problem that the price of used cars is difficult to estimate.