Research On On Personalized Recommend Service Of University Library Books Based On Spark Platform

In the era of big data and artificial intelligence,the rapid development of modern Internet generates large amounts of data every day.It is of great research significance how to mine massive data and extract valuable information.Big data platform technology research and development is the rise in this context.The era of big data is extremely rich in information and information,the problem of information overload is becoming more and more serious.The personalized recommendation system has become an important filtering tool to solve the problem of information overload and can help users to obtain the information they need in the massive data set.As an important gathering place for information,university libraries have a very large number of information resources.How to help readers find the books they are interested in has become an important research object in the field of personalized recommendation technology.In order to improve the efficiency of recommendation and solve the problem of system scalability,the recommendation system is built on the Spark platform.Using Spark's powerful big data processing capabilities,the performance of the recommendation system will be greatly improved.This paper is based on the Spark platform to research and implement personalized recommendation services for the library recommendation area of university libraries.The main research contents are as follows:(1)Read relevant literature at home and abroad,summarize the research status of personalized recommendation technology and personalized recommendation technology in the field of libraries,and study the origin,development situation,and basic hierarchical structure of big data technology and personalized recommendation technology.At the same time,a comparative analysis of the commonly used recommended technologies was conducted.(2)Spark platform optimization.During the parallelization of the recommendation algorithm using Spark,it was found that the Spark Shuffle Tasks caused memory wastage and the Spark operation efficiency was low due to unbalanced memory requirements.In order to solve the above problems,an adaptive Spark Shuffle Task memory scheduling algorithmbased on the improved fair scheduling algorithm is proposed in this paper.It is proved through experiments that the improved adaptive memory scheduling algorithm compares the default fair scheduling algorithm and reduces the Spark Task.Running time increases the overall performance of the cluster by 21.5%,thus ensuring the effective implementation of the upper layer algorithm.(3)Realization and Optimization of Book Recommendation Algorithm Based on Spark Platform.Using the optimized Spark platform,combined with the actual situation of university libraries,This article studies and implements collaborative filter recommendation based on reader-reader,based on the reader-book title recommendation.At the same time,for the problem of reader-book title recommendation results are single,a hybrid recommendation algorithm was proposed based on reader-book title recommendation.Based on the reader-book title recommendation algorithm and the reader-book classification collaborative filtering recommendation algorithm are combined,and the similarity calculation has been improved to increase the weight ? coefficient so that the similarity calculation can be dynamic with the actual situation.Change the size to improve the accuracy of similarity calculations.(4)Using the actual borrowing data of the school library from 2013 to 2017 to design relevant experiments to conduct empirical research.The final results show that the accuracy of the proposed hybrid recommendation algorithm is 90.8%,recall rate is 60.8%,and it is better than a single content-based recommendation algorithm.At the same time,a simple personalized book recommendation system based on Spark platform is designed,which mainly includes the overall framework of the system,the main modules of the system,and the reader's personal information,historical borrowing records,and personalized book recommendation list are realized in the form of front-end interface.,the borrowing of similar readers,and the borrowing of readers who have read this book,and other system function modules.The main contributions and improvements of this article are as follows:(1)In terms of Spark platform optimization,a memory scheduling algorithm based on fair scheduling algorithm improved self-adaptation was proposed to solve the problem that the memory was wasted due to the unbalanced memory requirements of each task and theSpark Shuffle operation efficiency was low,ensuring the upper layer.Recommend the effective implementation of the algorithm.(2)In the research of recommendation algorithms for books,a hybrid recommendation algorithm is proposed,and improvements have been made in the calculation of similarity.Weight ? coefficients have been added so that the similarity calculation can be dynamically changed in size with actual conditions and similarity calculations are improved.The accuracy improves the accuracy and recall of the recommendation algorithm.In the end,the accuracy of the proposed hybrid recommendation algorithm is 90.8%,and the recall rate is 60.8%,which is better than a single content-based recommendation algorithm.