Better online recommendations
Systems used for online shopping commonly make suggestions to users about other products. Yet the recommendations are often not fully relevant, necessitating a system that utilises all available customer information to make more targeted suggestions. The EU-funded 'Context-aware recommender systems (CARS)' (CARS) project aimed to build such a recommendation system for mobile and desktop devices. The system takes advantage of the full contextual relationship between user and item. The consortium's activities began at the end of 2011 and lasted two years. Achievements included developing five novel algorithms for collaborative filtering, which outperform competitors. The algorithms were integrated into an application (app) for Android mobile devices, called Frappe, which recommends other apps the user might like. The app uses contextual information from the mobile phone sensors to find similar products that may meet the apparent needs of the user. The project evaluated the app's usefulness and impact. Project results were published as conference and workshop papers, two of which received Best Paper awards. The CARS project advanced the field of context-based recommendation systems. As a result, new systems will provide more relevant suggestions, potentially leading to further purchases.
Keywords
Online retail, contextual relationship algorithms, customer information, targeted suggestions, context-aware, recommender system, Android, app, Frappe, mobile phone sensors
