Online Recommender Systems based on Data Stream Management Systems

Abstract

In this paper, I present a novel approach for implementing a stream-based Recommender System (RecSys). I propose to add RecSys operators to an application-independent Data Stream Management System (DSMS) to allow writing continuous queries over data streams that calculate personalized sets of recommendations. That empowers RecSys providers to create a custom RecSys by writing queries in a declarative query language. This approach ensures a flexible and extendable usage of RecSys functions in different settings and benefits from matured features of DSMSs.