Research And Design Of Context-aware Recommendation System Based On Hadoop

With the rapid development of Internet and IT, as well as cloud objects move wisdom(cloud computing, Internet, mobile Internet, big data and wisdom city) the emergence of new technologies, such as data growth and accumulation are very quickly, so as to cause "information overload" problem. Recommendation system as a branch of "personalized service", can effectively solve the problem. However, traditional recommendation system less concerned information such as context, and recommend few research at home and abroad about the context, thus this article focuses on context-aware recommendation system. At the same time, the puzzled people for years of large data storage and parallel computing problems, following the Hadoop distributed computing framework, effective solves this problem. Thus, context awareness recommendation system combined with Hadoop is the research subject.The main research work of this article is:(1) The first learn Hadoop, HBase, Hive, Sqoop, Flume, Storm and Kafka Hadoop series such as technology, and the recommendation system, pervasive computing, context-aware and a series of concepts;(2) Focus on content-based and context-aware recommendation algorithm, first described based content recommendation process in three steps to complete, and then modeling them every step and explain in detail the calculation of interest-vector model, and Applications "Bayes’ formula" to balance the problem of uneven distribution of data, then use cosine similarity to recommend content based formula, then the focus of the recommendation algorithm based on content, Hadoop and context information together, using MapReduce-based K-Means dimensionality reduction algorithm clustering. Finally, multidimensional scoring function, the recommended filter the results of contextual information;(3)Details of the Hadoop-based off-line context-aware recommendation system design work, from data acquisition to data processing, data storage, and then generate recommendations The whole process. Unlike other systems do not pay attention to the recommended data collection, data collection paper made a detailed design, but also highlights the ETL process Hive development and scheduling, and database design using HBase;(4) A brief introduction to the real-time recommendation framework of the system, and the use of technology;(5) Finally, experimental analysis and assessment, indicating plus Hadoop and context-aware content recommendation system information in a precise, recall and speedup on a clear advantage.