An Integration Of User Interest And Text Clustering For Segmenting Enterprises’ Customers In Microblogging

Based on the social network of users’ relationship, Microblog is an interactive communicating platform for information sharing and propagation. With the sharing, instantaneity, interactivity, influence of information propagation and the diversity of propagation mode, Microblog has not only become an important information medium for people to obtain hot news and various viewpoints, but also a social interaction platform for people to expand their interpersonal circles; And at the meantime, it has been a significant social media which cannot be ignored for enterprises to conduct precise marketing. The rapid development of Microblog has profoundly changed the fashion of interactive communication between enterprises and customers. Microblog have the characteristics of fast propagation, strong interactivity and real-time sharing, which can be fully exploited to conduct social marketing to improve enterprises’ brand and popularity for extending enterprises’ volume growth and profit gain. All of these turn microblog marketing into an important tool in enterprises’ social marketing.Customer segmentation is the vital fundamental basis to conduct precise microblog marketing. The most significant problem in customer segmentation is how to effectively represent customers’ characteristics. By far, there is little research about customer segmentation in social media home and board. Related research mainly focused on traditional methods towards customer segmentation, describing users’ characteristics in order to classify microblog users based on their demographic information and behavior messages in microblogging. However, due to a lack of comprehensive description to users’ interests and more detailed data about demographic information as well as behavior messages, traditional methods failed to achieve successful effects in customer segmentation, limiting their applications in customer segmentation in social marketing field.In this research, we consider that personal characteristics and social relationship are of great significance in describing customers’ characteristics which can offer a bran-new insight for customer segmentation in microblogging. Therefore, we propose an approach to generate textual descriptions of customers’ interests from their personal and social relationship perspectives by combining users’ self-defined tags, topical words extracted from micro texts and tags of the verified users they followed in certain fields. Moreover, based on text clustering techniques, we propose a framework to deal with description of customers’ interests, segmentation, evaluation and visualization in customer segmentation in microblogging. Experimental results indicate that the proposed framework is feasible and further customer segmentation method based on non-negative matrix factorization outperform that with K-means or hierarchical clustering methods.In full virtue of abundant valuable text contents and social relationship in microblogging, it will effectively represent customers’ interests combining users’ generated tags, topical words from micro texts and tags from the verified users they follow in certain fields. The proposed framework, which is an integration of user interest modeling and text clustering for customer segmentation, will not only provide a new research insight for customer segmentation in microblogging, but also present valuable references to some classical customer relationship management problems such as potential customer mining and customer churn prediction as well as some social marketing strategies such as personalized recommendation and precise advertising.