Optimisation of cross border export e-commerce supply chain network based on machine learning and random programming

This paper proposes a collaborative optimisation method that integrates machine learning and stochastic programming to address the high demand uncertainty, complex logistics links, and rising operating costs faced by cross-border e-commerce supply chain networks for exports. Firstly, a dynamic demand forecasting model is constructed using random forest and XGBoost algorithm. Secondly, based on the predicted results, a two-stage stochastic programming model is established. In the first stage, the overseas warehouse location and basic inventory configuration are decided, and in the second stage, dynamic replenishment strategies are generated. Further introduce an improved sample average approximation (SAA) algorithm to solve the model, and design a multi-objective evaluation system to balance cost, timeliness, and service level indicators. Through actual enterprise data verification, it has been shown that this method reduces total costs by 14.7% compared to traditional deterministic models, and the demand forecasting error is controlled within 8.5%.