Construction and Application of a Big-Data-Driven Demand Forecasting Model for Fresh-Produce E-Commerce

Abstract

This paper addresses the prominent operational challenges faced by China's fresh-produce e-commerce sector, including high demand uncertainty, large product loss rates, and coarse inventory management. From a big-data-driven perspective, we construct a lightweight demand forecasting model tailored to small- and medium-sized fresh-produce e-commerce enterprises and propose a set of accompanying dynamic inventory optimization strategies. The paper first reviews the current state and persistent problems of demand forecasting in fresh e-commerce. It then designs a four-dimensional multi-source data system covering sales, environment, scenario, and supply, together with a three-layer "data-algorithm-application" model architecture. Three inventory optimization strategies are then proposed: dynamic safety stock with precise ordering, differentiated category management, and end-to-end data collaboration. Finally, with Hema (Freshippo) as a case, this paper analyzes how its demand-forecasting and inventory-management practices can inform improvements to the proposed framework. The analysis shows that a big-data-driven lightweight forecasting framework can effectively raise the forecast accuracy and inventory turnover of small- and medium-sized fresh e-commerce enterprises without significantly increasing their technical threshold, offering useful references for the industry's digital transformation.