Residual Value of Used Car Analysis and Prediction

Huayi Jing, Xinfeng Ye, Sathiamoorthy Manoharan

Abstract: Car leasing is one of the most important business segments for automakers. The residual value plays a predominant role in determining the monthly installment in a lease agreement. Precision in forecasting a car's future residual value holds paramount importance for leasing companies. This study delves into the exploration of machine learning methodologies for the prediction of residual value. The study is based on real resale records of over 7,800 off-lease cars collected from one of the largest financial service companies in New Zealand. Additionally, macroeconomic variables such as GDP, CPI, interest rate, and used car index are included in the modeling process. We construct and assess the effectiveness of seven machine learning models. Our experimentation reveals that our CatBoost-based model attains a remarkable accuracy. It also halves the prediction error of the existing approach.