Hybrid Machine Learning for Used Car Price Determinants

This study develops a methodological framework for used car valuation through comparative analysis of regression techniques. Initial factor significance evaluation was conducted using Multiple Linear Regression (MLR) and Logistic Regression (LR), identifying ten key pricing determinants. Twelve regression models were subsequently implemented with these predictors, revealing LR's superior performance over MLR in coefficient stability and interpretability. The predictive phase comparatively evaluates Gradient Descent (GD) and Support Vector Machine (SVM) algorithms, with systematic residual analysis demonstrating SVM's optimal prediction accuracy. The findings provide empirical guidance for machine learning applications in automotive residual value estimation.