Vehicle Price Forecasting Based on Multiple Machine Learning Models

Abstract

Car consumers are paying greater attention to vehicle kinds as a result of the Internet’s rapid development and increasing transparency of vehicle configuration details. At the same time, dealers also want to predict the best-selling models and their sales according to the basic situation of car buyers. We use an open data set of Geely Automobile, which describes the information of 2,289 car buyers. Second, we anticipate the price of the car using the K-nearest neighbor model, the random forest model, LSTM (Long Short-Term Memory) model, among others. The experimental results show that our modified LSTM model outperforms other evaluation indices, such as ACC, RMSE, MSE, and MAE (Mean Absolute Error) (Accuracy). We hope that these data mining results can help car companies arrange production and help car buyers make decisions.