Data-driven analysis on anticipatory shipping for pickup point inventory

Abstract

The pickup point's inventory level is important for the online retailers who providing the same-day pickup service. The previous inventory optimization research of pickup point does not use the real-world transaction data or consider anticipatory shipping with emergency shipment strategy. In this study, we propose a forecasting-optimization integrated approach, “Machine learning - Quantile Regression”, to optimize pickup points anticipatory shipping inventory under considering emergency shipment based on the historical transaction data of online retailer. Compared with the original machine learning algorithms, “Machine learning - Quantile Regressio” can effectively increase the profits of online retailers, such as LGBM-QR, ANN-QR and LSTM-QR will respectively improve the profit 2.6%, 6.4% and 1.8% compared with LGBM, ANN and LSTM. We make interesting contributions: (i) we propose a data-driven solution to optimize anticipatory shipping inventories for online retailers under considering emergency shipment. (ii) we propose a novel algorithm LSTM-QR for anticipatory shipping inventory and demonstrate it outperforms other two algorithms.