基于云计算平台的图书馆混合推荐技术研究  

Research on library hybrid recommendation technology based on cloud computing platform

作　　者：赵冉[1] ZHAO Ran(Hebei Finance University,Baoding 071000,China)

机构地区：[1]河北金融学院

出　　处：《现代电子技术》2019年第23期145-149,共5页Modern Electronics Technique

摘　　要：研究基于云计算平台的图书馆混合推荐技术,向不同类型用户精准推荐图书。构建以并行计算框架为计算前提的图书馆混合推荐系统。采用Flume采集图书馆用户日志信息和图书馆图书数据,将采集到的数据信息传输到Kafka消息中间件后,向集群中传入集群订阅的数据信息,集群在用户粘稠度较低时采用聚类算法聚类图书馆数据信息,获取聚类中心,依照聚类中心用户读书偏好向目标用户推荐图书。在用户粘稠度较高时,采用基于协同过滤的混合推荐算法分析目标用户阅读偏好,计算目标用户偏好和用户群中用户偏好相似程度,将偏好相似程度作为权值,组合多个权值产生预测评分,通过预测评分推荐图书。经过实验分析发现,该方法推荐精准性最高为99.72%,推荐查准率高于95%,响应时间不超过0.2 s,说明该方法能高效快速地向目标用户推荐图书。The library hybrid recommendation technology based on cloud computing platform is studied in this paper to recommend books precisely to different types of users. A library hybrid recommendation system taking parallel computing framework as calculation precondition is constructed. Flume is used to collect library user log information and library book data. The collected data information is transmitted to Kafka message middleware,and then the cluster subscribing data information is transmitted to the cluster. When the user consistency is low,the cluster uses clustering algorithm to cluster the library data information,obtain clustering centers,and recommend books to target users according to the reading preferences of users in clustering center. When the consistency of users is high,a hybrid recommendation algorithm based on collaborative filtering is adopted to analyze target users′ reading preferences and calculate the similarity between target users′ preferences and user preferences in user groups.The preference similarity is taken as weight to combine multiple weights to generate predictive scores. Books are recommended according to the predictive scores. The experimental analysis results show this method′ s recommendation accuracy can reach to99.72%,the recommendation precision ratio is higher than 95%,and the response time is no more than 0.2 s,which indicate that the method can recommend books to target users efficiently and quickly.

关 键 词：云计算平台 图书馆 混合推荐 图书数据采集 数据信息聚类 协同过滤 

分 类 号：TN99-34[电子电信—信号与信息处理] TP302[电子电信—信息与通信工程]