The UbiCARS model-driven framework: Automating development of recommender systems for commerce

Abstract

Recommendations of products to customers are proved to boost sales, increase customer satisfaction and improve user experience, making recommender systems an important tool for retail businesses. With recent technological advancements in AmI and Ubiquitous Computing, the benefits of recommender systems can be enjoyed not only in e-commerce, but in the physical store scenario as well. However, developing effective context-aware recommender systems by non-expert practitioners is not an easy task due to the complexity of building the necessary data models and selecting and configuring recommendation algorithms. In this paper we apply the Model Driven Development paradigm on the physical commerce recommendation domain by defining a UbiCARS Domain Specific Modelling Language, a modelling editor and a system, that aim to reduce complexity, abstract the technical details and expedite the development and application of State-of-the-Art recommender systems in ubiquitous environments (physical retail stores), as well as to enable practitioners to utilize additional data resulting from ubiquitous user-product interaction in the recommendation process to improve recommendation accuracy.