Outlier identification and market segmentation using kernel-based clustering techniques

Customer relationship management (CRM) has become a major business strategy of the leading millennium since it can help decision makers to understand customers' profiles more clearly. For a successful CRM, it is important for a company to target the most profitable customers and to manage them through providing a variety of attractive and personalized goods or service. With proper market segmentation, companies can deploy the right resource to target groups and develop closer relationships with them more efficiently and effectively. Recently, there are many ways proposed by CRM researchers or marketers for effective market segmentation. Most of them, however, are not robust to outliers and/or cannot work well when the target clusters are overlapped. To solve this challenging problem, this paper using kernel-based clustering techniques presents a hybrid approach for outlier identification and robust segmentation in real application. Two real datasets, including the Iris and the automobile maintenance, are used to validate the proposed approach. Experimental results show that the proposed approach cannot only identify outliers in advance, but also achieve better segmentation.