基于Stacking模型融合的用户购买行为预测研究

在大数据时代背景下,如何利用大量的销售数据精准预测顾客未来需求,成为企业制定客户管理和库存管理决策的一个重要问题。目前关于用户购买行为预测的研究中很少能够预测用户具体的购买时间。基于已有的销售数据,提出了基于机器学习和Stacking集成的综合预测模型预测用户的购买行为,即未来是否购买及其购买时间。将模型应用在一家大型连锁零售企业的需求预测中,并对方法的有效性进行评估。结果表明,基于Stacking集成的融合模型对预测用户未来是否购买具有最佳性能,准确率达85%,AUC值达到0.928;LightGBM集成算法在预测用户购买时间时具有最优性能,相比于融合模型提升了5.5%的预测性能;融合模型+LightGBM算法的组合相比于均使用融合模型提升了9.4%的预测性能。

Stacking Fusion Model for Customer Purchase Behavior Prediction

In the era of the big data,how to use a large amount of sales data to accurately predict customers'future demand is an important issue for companies to make customer management and inventory management decisions.Currently,few studies on the prediction of consumers purchase behavior can predict the specific purchase time.Based on the existing sales data,this paper proposes a comprehensive prediction model based on the integration of machine learning and Stacking to predict future purchase behavior of consumers.We applied the model to the demand forecast of a large retail chain and evaluated the effectiveness of the method.The results show that the fusion model based on Stacking has the best performance for predicting whether consumers will purchase in the future,and the accuracy rate is 85%,the AUC value is 0.928;the LightGBM integrated algorithm has the best performance in predicting the consumer purchase time,which improves the prediction performance by 5.5%compared with the fusion model;the combination of the fusion model+LightGBM algorithm improves the prediction performance by 9.4%compared with that applying the fusion models in predicting both whether to buy and when to buy.