Abstract: The massive data about customer’s online shopping behavior contains a large amount of knowledge which can reflect the rules of customer’s online shopping. The knowledge is the important evidence of customer’s online shopping intention prediction and commodity recommendation. A customer purchase intention prediction system is discussed. The system is based on Hadoop and realizes the purchase intention prediction of online customers by using the knowledge obtained from the historical transaction data. On the basis of the knowledge base which is built with knowledge graph, statistical method and FP-Growth algorithm are leveraged to discover the knowledge from the transaction data in the way of incremental, and then makes the knowledge fusion into the knowledge base. The system can use the knowledge stored in the knowledge base to analyze and predict the purchase intention of online shopping customers thereby realize the goal that improves the operational efficiency of the e-commerce platform and the customer online shopping experience.

Key words: knowledge graph, customer’s online shopping intention prediction, MapReduce, knowledge update, graph retrieving

摘要： 电子商务网站所产生的海量客户网购行为数据中蕴含着丰富的反映客户网购行为规律的知识。这些知识是进行客户网购意向预测以及在此基础上进一步进行商品推荐的重要依据。基于Hadoop框架构建了“客户购买意向预测系统”，实现了利用隐含在交易数据中的知识对正在浏览中的电商客户的购买意向预测。在基于知识图谱技术构建网购客户购买行为知识库的基础上，利用FP-Growth算法和数据统计方法实现了以增量方式从购买者的交易数据中获取反映用户购买规律的知识，并将取得的知识融合到知识库中。系统能够利用存储于知识库中的知识在客户进行网上购物的过程中对其购买意向进行分析和预测，从而达到提高电子商务平台的运营效率，改善客户购物体验的目的。

关键词: 知识库, 网购意向预测, MapReduce, 知识更新, 图检索