Constructing stereotypes for an adaptive e-shop using AIN-based clustering

Abstract

This paper describes an adaptive electronic video store application that monitors customers' actions and provides dynamic movie recommendation. The adaptive recommendation is formed based on double stereotypes that have been constructed for user modeling. The construction of stereotypes has been based on a novel approach that uses an Immune Network System (INS). In particular, the INS has been applied on data collected from 150 users of an earlier version of the e-commerce application. Specifically, the INS clustered users' interests as well as movies and represented each resulting cluster with corresponding antibodies. The double classification (users' interests - movies) was performed in a hierarchical way that resulted in several levels of user stereotypes: These stereotypes are then used dynamically by the e-commerce application to infer users' interests in movies based on a small set of observed users' actions. © Springer-Verlag Berlin Heidelberg 2007.