Font Size: a A A

Research On Distributed Collaborative Filtering Recommendation Technology For Search Application

Posted on:2018-03-27Degree:MasterType:Thesis
Country:ChinaCandidate:H J ChenFull Text:PDF
GTID:2348330533456561Subject:Computer technology
Abstract/Summary:PDF Full Text Request
The rapid development of the Internet,making the scale of information access and the spread reached an unprecedented level,with the information overload,from the massive data to find the increasingly difficult.The emergence of information retrieval technology to alleviate this problem,and for different application scenarios,user needs are not clear or can not refine the demand for keywords,information retrieval technology appears to be powerless,which recommended technology came into being.Depth study of search technology principles and recommended mechanism,the effective combination of the two,simplify the recommendation process,optimize the recommended search accuracy.The main work of this paper is as follows:(1)Design a collaborative recommendation system framework for search applications: Including both offline and online;among them,the offline part is mainly composed of the recommendation engine and the retrieval database,the online part integrates the search service by combining the search database and provides the recommended model for the packaging application.(2)Research on three typical cooperative filtering algorithms: Including the userbased,Item-based and Slope-One algorithms in mahout;Through the experimental evaluation,Slope-One recommended algorithm which is based on on Item-based algorithm optimization recommended results are accurate and efficient.It is relatively simple to implement.(3)Implement a distributed collaborative recommendation engine: Including the Item-based algorithm based on the MapReduce calculation framework to optimize mahout;Through the optimization of the index model,to achieve similar computing based on the film data source parallel processing to improve the recommended efficiency,And through comparative experiments to analyze the data processing capabilities of the engine in stand-alone and distributed environments.(4)Implement recommended services based on the application of electricity providers: Including the recommended model of the package build and the combination of ElasticSearch search engine applications;And the decoupling and separation between the modules that are directly exchanged with the underlying data of the quotient platform,without affecting the normal function of other sites,join the recommended system module.Combine with the search engine ElasticSearch,a reasonable optimization of search field identifiers,improve the recommended process,reduce the recommended error.
Keywords/Search Tags:Collaborative filtering recommendation algorithm, Distributed Computing, MapReduce, HDFS, ElasticSearch
PDF Full Text Request
Related items