基于SOM神经网络的高校图书馆个性化推荐服务系统构建

The Recommendation System of University Library Personalized Service Based on SOM

资源分类不合理、资源检索机制不完善是高校图书馆数字化建设中的重要问题。文章基于SO M神经网络聚类算法无参数、精准度高和客观性强的特点,首先对山西大学图书馆用户Web访问行为进行聚类和优化分析。其次,基于用户分析结果的输出,将用户个人特征信息、用户行为数据以及文献数据库等相关数据资源进行筛选整合,形成可靠性和可用性更高的关联数据集,并结合语义检索和属性值匹配等技术,构建高校图书馆用户个性化推荐服务系统。最后对系统进行有效性验证,实现了图书馆内部主题推荐、图书推荐和专家推荐三个子系统的协同。通过用户与文献资源特征的相关性计算,进一步识别用户的兴趣点和所在聚类集。

The problems now for digital construction in university libraries are improper resource classification and incomplete resource retrieval mechanism. Based on SOM neural network clustering algorithm which is characterized by no arguments, high precision and strong objectivity, this paper first conducts the cluster and optimization analysis on the web access behavior of Shanxi university library users. The progress of clustering is divided into two stages： the rough adjustment training and the micro adjustment training, which can improve the clustering rate and effect. Based on the analysis results, this paper screens and integrates the data related to individual characteristics, behavior of users and literature database, which constitutes the linked data set with high reliability and availability. Combining with the semantic retrieval and attribute matching technology, a recommendation system of personalized service is established. Finally, the system is validated, which proves that it realizes the internal coordination among subject recommendation, book recommendation and expert recommendation. User interests and the clustering set are further identified by calculating the correlation between the characteristics of users and document resources.