Research on Used Car Transaction Price and Transaction Cycle Based on Model Fusion

Abstract:
In recent years, there has not been a reliable system for the transaction situation of the used car market in China. To solve the problem of pricing confusion in used car transactions, this paper uses historical transaction data to train the machine learning model: Bayesian Ridge, Elastic Network, Linear Regression, and BP neural network as the first-level model, and uses the Stacking model for evaluating training. In addition, we used stepwise regression to explore the influencing factors of the second-hand car transaction cycle. The results show that the fusion model has higher accuracy than the single model, and the last transaction adjustment interval, adjustments times, the deal price, and whether it is listed in the fourth quarter are the key factors affecting the transaction cycle.