基于GBDT的商品分配层次化预测模型
GBDT based hierarchical model for commodity distribution prediction

摘要 商品预测是使用以往商品信息去估计和推断未来商品的销售趋势，并以此作为对商品进行合理调配与规划的依据.为实现对商品销售的精确预测，在GBDT基础上，提出了一种层次化集成预测模型（HGBDT）.针对数据表征的高维问题，基于Bagging思想，在特征空间构建了该模型，实现对商品的有效描述，以此提高预测模型的性能与泛化能力.在开放数据库上的实验结果验证了本文模型的有效性.
Abstract：Commodity prediction uses the previous commodity information to estimate and infer the future trends of the commodity, and it can be used for carrying out reasonable planning and distribution of commodity. To achieve accurate forecast of merchandise sales, a commodity distribution prediction model (HGBDT) based on Gradient Boosting Decision Tree (GBDT)is proposed. To alleviate the problem of dimensionality curse, we construct a Bagging based hierarchical ensemble learning model.The temporal-spatial property of commodity is exploited for characterizing commodity effectively, which is beneficial to boost the generalization of the learned prediction model. Experimental results on open dataset demonstrate the effectiveness of the proposed method.