Research on used car price prediction based on random forest and LightGBM

Abstract

In recent years, while the scale of the used car market has been expanding, the price evaluation system of my country's second-hand car market has exposed the problem that it does not meet the market demand. Accurate used car price prediction can help people make correct decisions and avoid the wanton price tag of used cars in the market as much as possible. This paper uses the random forest and LightGBM algorithms to predict the price of used cars and compares and analyzes the prediction results. The experiments found that the relevant evaluation indicators of the random forest and LightGBM models are as follows: MSE is 0.0373 and 0.0385 respectively; MAE is 0.125 and 0.117 respectively; The R square of prediction is 0.936 and 0.933 respectively. Among the two prediction models, the prediction error of the LightGBM model is smaller, and it can be considered to be applied to other fields in future research.