An End-to-End Deep Learning Model for Solving Data-Driven Newsvendor Problem with Accessibility to Textual Word-of-Mouth Data

We study a data-driven single-period inventory management problem with uncertain demand, wherein a large amount of textual word-of-mouth and historical data are accessible. Unlike two-step frameworks (i.e., predict-then-optimization), we propose an end-to-end (E2E) framework that uses a deep learning model to directly output the suggested order quantity by inputting textual online reviews and other relevant feature data without any intermediate steps (e.g., text sentiment analysis). The E2E model does not require any prior assumptions about demand distribution and can automatically obtain the order quantity that minimizes newsvendor cost through the information from real-world data. We used public real-world data for the experiments. Compared with other data-driven models proposed in recent years, this model can significantly reduce the sum of overage and underage costs. Specifically, the addition of textual online review data improved ordering decisions by 31.8% cost reduction