تقسیم بندی مشتریان توسط داده معامله با سلسله مراتب مفهوم

The segmentation of customers is crucial for an organization wishing to develop appropriate promotion strategies for different clusters. Clustering customers provides an in-depth understanding of their behavior. However, previous studies have paid little attention to the similarity of different items in transaction. Lack of categories and concept levels of items, results from item-based segmentation methods are not as good as expected. Through employing a concept hierarchy of items, this study proposes a segmentation methodology to identify similarities between customers. First, the dissimilarity between transaction sequences is defined. Second, we adopt hierarchical clustering method to segment customers by their transaction data with concept hierarchy of consumed items. After segmentation, three cluster validation indices are used for optimizing the number of clusters of customers. Through the compassion of normalized index, the segmentation method proposed by this study rendered better results than other traditional methods.

Knowledge regarding what customers think, what they want, and how to serve them is quite useful for companies wishing to generate suitable strategies in competitive markets. Owing to disparate desires, interests, and needs, gaining a comprehensive understanding of customers is difficult. Since an organization cannot normally serve all customers in a market (Dibb & Stern, 1995), customer segmentation is often used by organizations to categorize customers for marketing purposes. Customer segmentation divides customers into groups, with the members of each group having similar needs, characteristics, or behaviors. Segmentation also represents the key element of customer identification in customer relationship management (Ngai, Xiu, & Chau, 2009). After segmenting customers, companies can then use further strategies such as customer attraction to maintain relationships with customers and gain more profit from them. The selection of the customers’ attributes is critical in their segmentation. The attributes for segmentation can be classified into two types: general attributes and transaction-based attributes (Tsai & Chiu, 2004). General attributes include customer base variables such as customer demographics, lifestyle, attitude and psychology (Bloom, 2005, Huang et al., 2007, Kuo et al., 2002, Lee and Park, 2005 and Vellido et al., 1999). The main goal of these studies is to offer appropriate services or products to people based on different customer status. This approach is known as customer status orientation. Although general attributes are easy to operate and understand, the disadvantage of using general attributes is that customers with similar general attributes do not necessarily have similar purchasing behavior, and information about customer variables is also difficult to collect and is often incomplete (Tsai & Chiu, 2004). Finally, the segmenting results obtained using general attributes may miss some important trends because of its static nature (Böttcher, Spott, Nauck, & Kruse, 2009). Articles using transaction-based attributes are mostly customer-value oriented (Böttcher et al., 2009, Chen et al., 2005, Cheng and Chen, 2009 and Hosseini et al., 2010). These articles focus on high value customers. Several articles have been made to take RFM (Hughes, 1994) as their clustering or mining attributes. Verhoef and Donkers (2001) used socio-demographic information and transaction information to measure customers’ potential value, and argued that companies should pay close attention to potentially valuable customers. Hwang, Jung, and Suh (2004) proposed a new lifetime value model by considering past profit contribution, potential benefit, and defection probability of customers, and finally segmenting customers based on their value. Furthermore, Kim, Jung, Suh, and Hwang (2006) used a lifetime value model with current value, potential value, and customer loyalty in segmentation. They also mentioned that current value provides a financial viewpoint, potential value indicates cross-selling opportunities, and customer loyalty estimates durability of the previous two values. These articles focus on customer value in order to gain the most profits for firms. However, from a customer retention standpoint, these approaches may be not good enough to ensure firm/customer relationship are maintained in the long-term, because firms do not have any idea what customers like or prefer, and the approaches lack the most important information about products. There are only a few articles dealing with transaction-based attributes which consider product information as an important factor in segmentation (Lu and Wu, 2009, Tsai and Chiu, 2004 and Tsai and Shieh, 2009). Because the studies by Lu and Wu, 2009, Tsai and Chiu, 2004 and Tsai and Shieh, 2009 looked at product information, they can be classified as customer preference-oriented. By discovering customers’ preferences, firms can then deliver the right marketing strategy to the right customer cluster, and ultimately can improve the quality of the customer relationship and enhance customer loyalty. Although these three studies considered product information, they did not specifically consider the relationships among items. When there are huge numbers of items provided by an enterprise, this means that similarities between any two customers are often actually very small, owing to customer preferences for similar, but not exactly the same, items. Therefore, segmentation results in these studies have not been as good as expected. Based on a concept hierarchy of items, this study proposed a segmentation methodology to identify relationships between customers. Generally speaking, two more similar items have strong relationship than that between two less ones. The rest of this paper is organized as follows: Section 2 describes an overview of the related research including data representation, similarity measure and its example, hierarchical clustering algorithm, and clustering criteria functions. Section 3 presents the proposed procedure and briefly discusses its architecture. Section 4 analyzes the experimental results. Finally, a summary and conclusion are presented in Section 5.

Most customer preference-oriented segmentation methods described in Section 1 consider the same-or-not instead of relationship among items. This kind of approach renders a critical loss of information. In this study, we take the concept hierarchy of items into consideration for segmenting customer through their transactions. First, the dissimilarity between items is employeed, and then the dissimilarity between transaction sequences is employeed. Next, hierarchical clustering methods are employeed to derive the clustering results. Following this, the best clustering result, which depends on a average index, are obtained. Finally, the visualization of results can be consulted to facilitate decision making. From the implementation results, the proposed method in this study outperforms the traditional method. Besides, a mini system is implemented by this study. From this system, manager in bookstore or supermarket can realize the clusters of customers and their preference of products. This information is quite useful for market segmentation and product promotion. Further research may consider the sequence of transactions for time-serial analysis.