SARI OpenRec -- Empowering Recommendation Systems with Business Events

Abstract

With growing product portfolios of eCommerce companies it gets increasingly challenging for customers to find the products they like best. Current recommendation approaches primarily rely on customer-product affinities derived from explicit ratings or historical purchases. In this paper, we introduce SARI OpenRec, an extendible framework combining the capabilities of complex event processing and recommendation systems. SARI OpenRec enhances recommendations by considering most recent customer activities reflected in event streams. These include website activities (e.g., page views, advertisement clicks, page durations), as well as business activities such as purchases, payments and returned goods. The integrated rule engine enables companies to model rules for dynamically adjusting the recommendation based on stock levels, seasonal factors or current marketing campaigns. Finally, we demonstrate how to analyze historic events and evaluate the recommendation process using the visualization facilities in SARI OpenRec. We claim that by considering a wide range of external signals and business events, the recommendation system becomes more context-aware and personalized.