Data-driven optimization models for inventory and financing decisions in online retailing platforms

With data-driven optimization, this study investigates the sellers' inventory replenishment and financial decisions, and lenders' interest rate decisions in online retailing platforms. Moreover, we focus on the annual large-scale promotion, which requires massive capital in a short period. While scholars studying the data-driven inventory replenishment problem hardly consider capital-constrained sellers, these problems are important because the seller's capital level can significantly influence the order quantity and generate different effects on inventory management. Hence, we propose two novel data-driven game-theoretic approaches (including separated and integrated methods) using machine learning and deep learning methods to optimize inventory replenishment and financial decisions for the sellers who obtain financial support from the online platform. Moreover, we propose a data-driven game-theoretic model for the online platform to optimize their interest rate considering the market potential. We explore the real retailing transaction data containing 199,390 weekly sales records. We find that the seller and lender can benefit when the seller chooses integrated machine learning and quantile regression method. However, we find that only a low capital level can motivate the seller to choose to borrow from the lender. Interestingly, our results also suggest that the lender has the motivation to build a data-driven system that helps sellers optimize inventory decisions. Our work identifies the optimal interest rate and inventory decision under the data-driven method. We propose data-driven decision support tools by evaluating the values of both the lender's and the seller's profit and provide new management insights on joint inventory and financing decisions.