基于背景的个性化客户行为模型研究

Research of context-based personalized customer behavior model

针对商业客户的背景信息对其消费行为的影响问题,提出了一种基于背景的商业客户行为模型的构建方法。该方法不仅收集了包含三级粒度背景信息的某大学学生客户的网上交易数据并按依赖于交易数据项的统计量对客户进行分组,还利用朴素贝叶斯(NB和TAN)及分组和统计关系数据库(GAC-RDB)分类器学习了各客户分组的背景和非背景预测函数,同时使用各预测变量的受试者运行特征曲线下面积(AUC)值,对客户背景在预测客户购买行为时的作用进行了定量的比较和分析。研究结果表明:背景信息对客户特别是个性化客户(单一客户)的消费决策具有良好的预测效果。

To investigate the context＇s influence on business customer consumption behavior, a sort of construction approach of context-based customer behavior model was proposed. One undergraduate customer transaction data online with three-level context granularity was collected and grouped on statistic based transaction data item. The classifiers including Naive Bayesian （NB）, Tree Augmented NB （TAN） and Grouping and Counting-relational Database （GAC-RDB） were used to learn context and non-context predicating functions of each customer group. Based on the Area under a Receiver Operating Characteristic Curve （AUC） of predicating variable, the paper compared and analyzed quantitatively the effect of customer context when predicating his buying behavior. The experimental results demonstrate that the context information has preferable predication performance on the consumption decision of the customers especially the personalized customers.