Second-hand Car Price Prediction Based on a Mixed-Weighted Regression Model

With the development of motor vehicles, the circulation demand of motor vehicles in the form of "second-hand cars" in circulation links is increasing. As a special "e-commerce commodity", second-hand cars are more complicated than ordinary e-commerce commodities. As a result, it is difficult to estimate the price of second-hand cars, which is not only influenced by the basic configuration of the car, but also by the car conditions. At present, the state has not issued a standard to judge the value of second-hand car. To solve this problem, in this paper, first making feature engineering, which includes data preprocessing and feature screening. Data preprocessing includes data cleaning and data transformation, data cleaning includes removing outliers and filling missing values, and data transformation is used to unify data format to improve data quality. The feature screening includes correlation analysis and feature extraction based on LightMBG, and the screened features provide the basis for model building, training and prediction. Then, five regression models are constructed by using the feature attributes obtained by the feature engineering for training, and evaluated. Then, Random Forest and XGBoost are weighted and mixed to got a novel regression model, and the effect of the model is better than that of the five regression models. Finally, the novel regression model is used to predict the price of second-hand cars.