基于内存计算的图书馆文献服务模式构建研究  

Research on the Construction of Library Literature Service Model Based on Memory Computing

作　　者：于芳[1] YU Fang(Library,Harbin Institute of Technology in Weihai,Weihai 264209,China)

机构地区：[1]哈尔滨工业大学(威海)图书馆,山东威海264209

出　　处：《微型电脑应用》2021年第5期48-51,共4页Microcomputer Applications

基　　金：山东省高等学校人文社会科学研究计划项目(j11w151)。

摘　　要：针对传统文献推荐过程中易于发生文献查找困难、文献浏览迷失等问题,基于大数据特征,利用内存计算中Spark系统框架高的容错机制和实时运算优势,提出了一种“混合关联”的图书馆推荐算法。利用Spark RDD来支撑“字符串匹配”,利用Spark MLlib支撑“相似度匹配”,通过TF-IDF()算法获得分词的TF/IDF值作为权重值,建立起文献、混合权重的Spark的三元组形式,并利用混合权重值排名建立不同长度推荐列表,以准确率对推荐算法的性能进行了评价,结果表明该算法在庞大图书系统中依然具备了非常高的文献推荐准确率,能够满足用户对感兴趣资料文献的查找需求。The traditional literature recommendation process is easy to occur literature search defects,literature browsing lost and so on.Based on characteristics of big data,a library recommendation algorithm of“mixed association”is proposed by using the high fault tolerance mechanism and real-time operation advantage of Spark system framework in memory computing.We use Spark RDD to support“string matching”,use Spark MLlib to support“similarity matching”,use TF-IDF()algorithm to obtain the TF/IDF value of participle as the weight value and establish the ternary form of Spark with literature and mixed weight.The weight ranking is used to establish the recommendation list with different lengths,and the performance of the recommendation algorithm is evaluated by accuracy.The results show that the algorithm has very high literature recommendation accuracy in the huge book system,which can meet the user's search demand for the information of interest.

关 键 词：内存算法 文献推荐 Spark框架 混合权重 

分 类 号：G250[文化科学—图书馆学]