A new data imputation technique for efficient used car price forecasting

Abstract

This research presents an innovative methodology for addressing missing data challenges, specifically applied to predicting the resale value of used vehicles. The study integrates a tailored feature selection algorithm with a sophisticated imputation strategy utilizing the HistGradientBoostingRegressor to enhance efficiency and accuracy while maintaining data fidelity. The approach effectively resolves data preprocessing and missing value imputation issues in complex datasets. A comprehensive flowchart delineates the process from initial data acquisition and integration to ultimate preprocessing steps, encompassing feature engineering, data partitioning, model training, and imputation procedures. The results demonstrate the superiority of the HistGradientBoostingRegressor for imputation over conventional methods, with boosted models eXtreme gradient boosting (XGBoost) regressor and gradient boosting regressor exhibiting exceptional performance in price forecasting. While the study’s potential limitations include generalizability across diverse datasets, its applications include enhancing pricing models in the automotive sector and improving data quality in large-scale market analyses.