| The valuation of used automobiles is a critical economic problem with significant implications for insurers, leasing companies, and individual consumers. While modern Machine Learning (ML) approaches offer superior predictive power, they often lack the interpretability required for trusted financial applications. This paper introduces the Dynamic Feature Depreciation Network (DFDN), a novel neural architecture designed to bridge the gap between econometric transparency and deep learning accuracy. DFDN utilizes a hybrid structure to explicitly model the heterogeneous depreciation rates of vehicle attributes. |
*** Title, author list and abstract as submitted during Camera-Ready version delivery. Small changes that may have occurred during processing by Springer may not appear in this window.