Implementation of Context-Aware Item Recommendation through MapReduce Data Aggregation

Abstract

As the amount of ubiquitous product and service information within our daily lives is exploding, client-centric and context-aware information filtering is one of the thriving topics within the next years. A popular approach is to combine context-awareness with traditional recommendation engines in order to evaluate the relevance of a large amount of items for a given situation and user. Within this work we propose a general software architecture as well as a prototypical implementation for a framework that combines traditional recommendation methods with a variable number of context dimensions, such as location of social context. This work shows how to use a MapReduce programming model for aggregating the necessary information for calculating fast context-aware recommendations. A use-case at the end of this work shows how to use this general framework to implement a client-centric, MapReduce-based recommendation engine for real-time recommending music events.