Customer classification models and recommender systems for ubiquitous customer relationship management

Abstract

As the Internet and mobile devices spread around the world, many people have become interested in customer relationship management (CRM) using Web or mobile technology. These technologies provide new opportunities for service providers to increase sales and customer satisfaction. However, in order to obtain these positive outcomes, companies should understand their customers in depth, and they should be able to predict customer behavioral patterns accurately. As a result, building an appropriate customer classification model and a recommendation model has become one of the most important issues in the analytic CRM area under the existing ubiquitous environment. Customer classification means that a company classifies its customers into predefined groups having similar behavior patterns. In general, companies construct a customer classification model to find the prospects for the specific product. This kind of knowledge may create a variety of marketing opportunities for the company such as one-to-one marketing, direct mail, and sales promotion via the telephone or e-mail. A recommendation model is used to help customers who are searching for products they would like to purchase by creating a list of recommended products. It is also used as a tool for accelerating cross-selling and strengthening customer loyalty. Both of these models require a sophisticated analytic process to provide beneficial and useful information for a company. However, there are many barriers that interfere with the process, thus researchers have focused on the issues surrounding the enhancement of the process to build these models. This study also deals with the topics of improving the performance - prediction accuracy and customer satisfaction - for the customer classification and recommendation models. We propose novel approaches for two areas to build the customer classification and recommendation models, (1) appropriate preprocessing techniques and (2) enhanced classification techn...