Abstract: Value segmentation and identifying important value customers is vital to airline profitability. In this paper, we develop the value segmentation model of airline customers based on K-means and neighborhood rough set. Firstly, according to the current value and potential value of customers, comprehensive value evaluation index system of airline customers is established. Next, customers are clustered using k-means based on Elbow method. Neighborhood rough set modelis used for attribute reduction. Then, customers are grouped by the reduced decision system. SMOTE method is used to remove imbalance of data before evaluation. Then, Grid search method is used to optimize parameters of ensemble classifiers and segmentation results are evaluated. At last, customer value segmentation is finished according to the evaluation results. In the end, 62988real airline customer records are used for empirical analysis. Theresults show that classification accuracy of the potential VIP customers reachesto92%. The model we propose provides a new idea for value segmentation of airline customers.

Key words: airline customers, value segmentation, K-means, neighborhood rough set, ensemble classifiers

摘要： 对客户价值进行分类, 识别重要价值客户, 对航空公司获利至关重要。本文提出了基于k-means和邻域粗糙集的航空客户价值分类模型。首先, 从客户的当前价值和潜在价值双视角出发, 建立了航空客户综合价值评价指标体系; 之后, 采用基于Elbow的k-means方法对航空客户进行聚类, 采用邻域粗糙集方法对决策系统进行指标约简, 根据约简后的决策系统完成客户价值初筛。评估前先使用SMOTE方法消除数据的不平衡性, 而后采用网格搜索组合分类器的方法对航空客户价值分类的效果进行评估和检验。最后, 根据评估结果对航空客户价值细分。文末, 对国内某航空公司的62988条真实客户记录进行了实证分析和验证, 其中, 潜在VIP客户群的分类准确率达到了92%, 从而为航空客户价值分类提供了一种新思路。

关键词: 航空客户, 价值分类, K-means, 邻域粗糙集, 组合分类器