Files

Abstract

Countries with a well-developed auto manufacturing industry, such as Germany and the United States, often have a well-structured secondary market for cars. There are credible third-party platforms, such as Kelly Blue Book (KBB), DAT and Schwack-Liste, to evaluate the price of used cars. Such platforms do not exist in China. In this paper, I use quantitative methods to establish a model for used car price prediction for car market in Shanghai, China. I compare LASSO regression, Generalized Linear Model, Decision Tree and Artificial Neural Network (ANN) models in order to determine the best prediction. Among them, our ANN model performed best, with model Mean Squared Error (MSE) of 2.11 and prediction MSE of 5.85. That is, given an out-of-sample car to be estimated, our ANN model can give a prediction price with a floating error of 58k CNY. While for the cars in the training set, the prediction has a more accurate estimate, with an error of 21k CNY. Although our model still needs further improvements, it is valuable for buyers and sellers of used cars to get a guidance price for their vehicles.

Details

Actions

from
to
Export
Download Full History