A Context-Aware Recommendation System Based on Hadoop Big Data

Muhammad Ayoob DARS, Liu QINGLING, Zhang CHAOZHU, Shahabuddin SHAIKH, Ghulam FARID

Abstract

The data growth rate is rapidly increasing with the emergence of new concepts and techniques such as Cloud Computing, Big Data, IoT and Mobile Cloud Computing. This generates the problem of Information Overload, and obtaining information becomes difficult for users. Simultaneously, the problems of parallel computing and big data storage have emerged, which has puzzled people for many years, and they have been adequately addressed after the emergence of the Hadoop distributed framework. As such, an in-depth investigation of Hadoop and context-aware recommendation systems was conducted and this paper demonstrated the connectivity between them. Further on, context-based and content-based recommendation algorithms were integrated using the MapReduce framework based on Hadoop. The K-Means algorithm performed clustering and dimensionality reduction and used a multi-dimensional scoring function to filter the recommendation results. The design of the offline context-aware recommendation system was addressed in detail, including the entire procedure for collecting information, processing data, and storing data. This paper also focused on the process of developing the Hive data warehouse tool and utilising it for ETL (Extract-Transform-Load) processing, and employed HBase for the design of the proposed system. The results of the conducted experimental analysis proved the necessity of integrating the Hadoop distributed framework and highlighted the advantages of the proposed content-based recommendation system using context-aware information with regard to accuracy, recall and the acceleration ratio in comparison with the traditional recommendation algorithms.

Keywords

Context-Aware, Content-Based, Hadoop, MapReduce, Recommendation System.

View full article