A conceptual framework for building a mobile services' recommendation engine

This paper presents a conceptual framework for building a recommendation engine for use in the ubiquitous consumer wireless world (UCWW). The framework is based on the Lambda Architecture and as such provides real-time recommendations at the speed layer and off-time analytical operations at the batch layer. Moreover, at the speed layer, a root server is used for updating the real-time recommendation algorithms and dispatching the processing of users' information to different recommendation servers for load balancing with a consistent hash algorithm. At the batch layer, a number of off-line machine learning algorithms are utilized for items analyzing, user profiling based on raw data, and sending the results to the serving layer. At the end, a decision server formats recommendations for ‘best’ service instances, which are sent back to users for effecting an always best connected and best served (ABC&S) experience.