Residual Value Evaluation of Operating Pure Electric Vehicles Based on Machine Learning

Abstract

In view of the current imperfect second-hand evaluation system for new energy vehicles, there are limitations such as simple evaluation models, strong subjectivity, and large evaluation differences. This paper establishes the residual value evaluation model for operating pure electric vehicles, combining actual vehicle operation and maintenance data and establishes the residual value rate evaluation model based on the XGBoost algorithm and Boosted Trees enhanced algorithm. The multi-dimensional feature data of vehicle type, service time, mileage and region are extracted by feature engineering, and the evaluation system of residual value rate correction coefficient is established based on AHP. Finally, based on residual value rate evaluation model and the residual value rate correction coefficient system model to optimize the replacement cost method evaluation model, thereby constructing a complete residual value evaluation model. The model is based on actual vehicle operating data and machine learning algorithms, which has strong pertinence and real-time for the residual value evaluation of pure electric vehicles.