Design And Implementation Of A Recommendation Algorithm For Large-scale Scientific Literature

We are in the age of big data, the number of various information data has become increasingly large, and how to quickly find the information they need from large-scale information is becoming increasingly difficult. Therefore, information retrieval and information recommendation technology has become increasingly important. For a search function of the system, if we can accurately and efficiently recommended information that users interested in, which will be largely improved user experience. But also further improve the system’s user evaluation, bring more users. Information recommendation technology for big data era is well worth studying.For researchers, students, scientific literature is very important research reference. Now face the increasingly large-scale scientific literature, how to help the scientific literature demander quickly and efficiently find the document from the mass of the electron scientific literature has become an urgent problem. How to rely on studying different users’ interests and scientific literature resource characteristics, to initiative recommend users interested literature for them is one of the key core issue of Scientific Literature Information Service System. In this paper, design and implement scientific literature recommended strategy of a large-scale scientific literature retrieval system. Recommended for scientific literature mainly do recommend specific research the strategies and algorithms in real-time recommendation and the timing lines off recommendation. Recommended real-time scientific literature focused on when users are searching and browsing literature, Combining Lucene search and based on content recommendation real-time recommend search terms and similar content and related literature author for the users. Real-time recommendation has better real-time Performance, similar content recommendation has high accuracy rate. Timing lined off recommend by emails sending for users to recommend literature may be users’ interest. to study how to analyze user behavior, set up user interest model also studied how to accurately characterize the literature resources, said Construction vector Model References. Based on the contents of different content weights of different user behavior weighting, whether added weight, the effect of different weights effect of the recommended experimental study. Experimental results show that the recommendation that based on the content of Document Resources weighted and behavior weighted can more accurate recommended for users. And modeling large-scale literature use of Hadoop MapReduce technology platform for distributed computing research. And to analyze the experimental results, which provides a basis for future large-scale recommendation research and computational efficiency.