ABSTRACT
Since the prosperity of mobile communication and Internet, E-commerce is booming and it has promoted the explosive growth of express parcel volume. At present, parcel delivery is faced with common problems for E-commerce companies, especially customers have higher requirements for timeliness. Considering that the parcel delivery industry is faced with the characteristics of diversified customer sources and consumer personalized demand, the multi-warehouse inventory can help E-commerce platforms effectively shorten the delivery time and reduce logistics costs by pre-allocating commodity inventory. This paper designs a prediction model of commodity demand based on customer behavior to support the strategy of “commodity first, order later to”. Then, it applies data exploration and feature engineering to preprocesses the massive historical order data stored by the e-commerce platform. Finally, the machine learning algorithm of GBDT is applied to analyze the historical order data, to predict and obtain the stock scheme of multi-warehouse for E-commerce platform.
