Personalization of newsletters using multistage fuzzy inference

Abstract

Personalization is becoming more and more important in any e-commerce application. However, nowadays it is primarily used to personalize web-pages whereas newsletters are in most cases static, i.e., independent of the customer's profile. In this paper we introduce a method for personalizing newsletters based on fuzzy IF-THEN rules. The key intention for using fuzzy rules (instead of crisp rules) is to provide a framework that can easily be used to express vague concepts as they are frequently used in marketing issues. In particular, fuzzy rules are easy to understand (and define) for human operators like marketing experts. Our approach makes use of multistage inference. This guarantees that the definition of the rulebase can be done in a modular fashion and hence the reuse of rules is directly supported. © Springer-Verlag 2001.